Models

What builders use, how models score, and what they cost. Usage comes from public Tokscale submissions. Scores come from Epoch AI, LMArena and Terminal-Bench, each with its own unit, date and run conditions. Prices are list prices from models.dev.

Columns

Sort by

Oct 10, 2024 – Nov 13, 2026 (UTC)2,457 accounts100% of tokens from an identified modelData through Nov 13, 2026, 01:09 UTCHow we count

944 models: the 944 with 5+ active users in this period.All models · About the benchmarks

Models by Active users, all time
Model		Top agent
Claude Haiku 4.51. Anthropic	1,807	Claude Code91%	142.8 Epoch Capabilities Index: 142.8 Model Claude Haiku 4.5 Listed as Claude Haiku 4.5 Conditions Not reported Interval 141.1–144.5 Unit ECI Model released Oct 15, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,330 WebDev Arena: 1,330 Model Claude Haiku 4.5 Listed as Claude Haiku 4.5 (20251001) Conditions Not reported Interval 1325–1335 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.19 Estimate at list prices Estimate $0.19 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $1, output $5, cache read $0.10, cache write $1.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.52. OpenAI	1,586	Codex CLI89%	160.1 Epoch Capabilities Index: 160.1 Model GPT-5.5 Listed as GPT-5.5 Conditions Not reported Interval 158.1–162.4 Unit ECI Model released Apr 23, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,513XHigh · Codex WebDev Arena: 1,513 Model GPT-5.5 Listed as GPT-5.5 (xHigh, Codex Harness) Conditions XHigh · Codex Interval 1507–1519 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 3 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.94 Estimate at list prices Estimate $0.94 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $5, output $30, cache read $0.50, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Claude Sonnet 4.63. Anthropic	1,566	Claude Code71%	152.9 Epoch Capabilities Index: 152.9 Model Claude Sonnet 4.6 Listed as Claude Sonnet 4.6 Conditions Not reported Interval 151.2–154.7 Unit ECI Model released Feb 17, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,523 WebDev Arena: 1,523 Model Claude Sonnet 4.6 Listed as Claude Sonnet 4.6 Conditions Not reported Interval 1518–1528 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.56 Estimate at list prices Estimate $0.56 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $3, output $15, cache read $0.30, cache write $3.75 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.44. OpenAI	1,529	Codex CLI88%	157.8 Epoch Capabilities Index: 157.8 Model GPT-5.4 Listed as GPT-5.4 Conditions Not reported Interval 155.9–159.9 Unit ECI Model released Mar 5, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,464High · Codex WebDev Arena: 1,464 Model GPT-5.4 Listed as GPT-5.4 (High, Codex Harness) Conditions High · Codex Interval 1445–1483 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 3 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.47 Estimate at list prices Estimate $0.47 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $2.50, output $15, cache read $0.25, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Claude Opus 4.65. Anthropic	1,414	Claude Code78%	156.3 Epoch Capabilities Index: 156.3 Model Claude Opus 4.6 Listed as Claude Opus 4.6 Conditions Not reported Interval 154.6–158.3 Unit ECI Model released Feb 5, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,547High WebDev Arena: 1,547 Model Claude Opus 4.6 Listed as Claude Opus 4.6 (High) Conditions High Interval 1541–1552 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.94 Estimate at list prices Estimate $0.94 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.3 Codex6. OpenAI	1,335	Codex CLI85%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	1,407Codex WebDev Arena: 1,407 Model GPT-5.3 Codex Listed as GPT-5.3-Codex (Codex Harness) Conditions Codex Interval 1393–1420 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.35 Estimate at list prices Estimate $0.35 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $1.75, output $14, cache read $0.17, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GPT-5.6 Sol7. OpenAI	1,262	Codex CLI85%	163.0 Epoch Capabilities Index: 163.0 Model GPT-5.6 Sol Listed as GPT-5.6 Sol Conditions Not reported Interval 160.7–165.5 Unit ECI Model released Jul 9, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	38.5%Max · Codex Terminal-Bench 4.1: 38.5% Model GPT-5.6 Sol Listed as GPT-5.6 Sol Conditions Max · Codex Interval 34.6–42.3% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,616XHigh · Codex WebDev Arena: 1,616 Model GPT-5.6 Sol Listed as GPT-5.6 Sol (xHigh, Codex Harness) Conditions XHigh · Codex Interval 1610–1622 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.75 Estimate at list prices Estimate $0.75 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $4, output $20, cache read $0.40, cache write $5 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Opus 4.88. Anthropic	1,258	Claude Code87%	159.9 Epoch Capabilities Index: 159.9 Model Claude Opus 4.8 Listed as Claude Opus 4.8 Conditions Not reported Interval 157.9–162.2 Unit ECI Model released May 28, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	23.6%Max · Claude Code Terminal-Bench 4.1: 23.6% Model Claude Opus 4.8 Listed as Opus 4.8 Conditions Max · Claude Code Interval 20.1–27.2% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,556High WebDev Arena: 1,556 Model Claude Opus 4.8 Listed as Claude Opus 4.8 (High) Conditions High Interval 1550–1562 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.94 Estimate at list prices Estimate $0.94 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.6 Terra9. OpenAI	1,106	Codex CLI71%	160.7 Epoch Capabilities Index: 160.7 Model GPT-5.6 Terra Listed as GPT-5.6 Terra Conditions Not reported Interval 158.5–163.3 Unit ECI Model released Jul 9, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	21.5%Max · Codex Terminal-Bench 4.1: 21.5% Model GPT-5.6 Terra Listed as GPT-5.6 Terra Conditions Max · Codex Interval 18.3–24.8% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,522XHigh · Codex WebDev Arena: 1,522 Model GPT-5.6 Terra Listed as GPT-5.6 Terra (xHigh, Codex Harness) Conditions XHigh · Codex Interval 1515–1528 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.39 Estimate at list prices Estimate $0.39 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $2, output $12, cache read $0.20, cache write $2.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.6 Luna10. OpenAI	1,079	Codex CLI74%	156.7 Epoch Capabilities Index: 156.7 Model GPT-5.6 Luna Listed as GPT-5.6 Luna Conditions Not reported Interval 154.8–158.7 Unit ECI Model released Jul 9, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	17.6%Max · Codex Terminal-Bench 4.1: 17.6% Model GPT-5.6 Luna Listed as GPT-5.6 Luna Conditions Max · Codex Interval 14.7–20.5% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,518XHigh · Codex WebDev Arena: 1,518 Model GPT-5.6 Luna Listed as GPT-5.6 Luna (xHigh, Codex Harness) Conditions XHigh · Codex Interval 1512–1525 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.04 Estimate at list prices Estimate $0.04 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $0.20, output $1.20, cache read $0.02, cache write $0.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Sonnet 511. Anthropic	1,056	Claude Code89%	156.9 Epoch Capabilities Index: 156.9 Model Claude Sonnet 5 Listed as Claude Sonnet 5 Conditions Not reported Interval 154.8–159.2 Unit ECI Model released Jun 30, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	12.1%Max · Claude Code Terminal-Bench 4.1: 12.1% Model Claude Sonnet 5 Listed as Sonnet 5 Conditions Max · Claude Code Interval 9.1–15.1% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,540High WebDev Arena: 1,540 Model Claude Sonnet 5 Listed as Claude Sonnet 5 (High) Conditions High Interval 1534–1546 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.37 Estimate at list prices Estimate $0.37 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $2, output $10, cache read $0.20, cache write $2.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Opus 512. Anthropic	1,050	Claude Code88%	164.7 Epoch Capabilities Index: 164.7 Model Claude Opus 5 Listed as Claude Opus 5 Conditions Not reported Interval 162.1–167.5 Unit ECI Model released Jul 24, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	53.3%XHigh · Claude Code Terminal-Bench 4.1: 53.3% Model Claude Opus 5 Listed as Opus 5 Conditions XHigh · Claude Code Interval 50.2–56.4% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Best of 5 reported results for this model. Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,689Max WebDev Arena: 1,689 Model Claude Opus 5 Listed as Claude Opus 5 (Max) Conditions Max Interval 1683–1695 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.94 Estimate at list prices Estimate $0.94 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Opus 4.713. Anthropic	1,043	Claude Code87%	157.4 Epoch Capabilities Index: 157.4 Model Claude Opus 4.7 Listed as Claude Opus 4.7 Conditions Not reported Interval 155.5–159.6 Unit ECI Model released Apr 16, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,558High WebDev Arena: 1,558 Model Claude Opus 4.7 Listed as Claude Opus 4.7 (High) Conditions High Interval 1552–1564 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.94 Estimate at list prices Estimate $0.94 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.4 Mini14. OpenAI	1,041	Codex CLI74%	149.3 Epoch Capabilities Index: 149.3 Model GPT-5.4 Mini Listed as GPT-5.4 Mini Conditions Not reported Interval 147.7–151.1 Unit ECI Model released Mar 17, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,397High WebDev Arena: 1,397 Model GPT-5.4 Mini Listed as GPT-5.4 mini (High) Conditions High Interval 1391–1404 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.14 Estimate at list prices Estimate $0.14 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $0.75, output $4.50, cache read $0.07, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Claude Fable 515. Anthropic	1,020	Claude Code87%	164.1 Epoch Capabilities Index: 164.1 Model Claude Fable 5 Listed as Claude Fable 5 Conditions Not reported Interval 161.7–166.8 Unit ECI Model released Jun 9, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	44.9%Max · Claude Code Terminal-Bench 4.1: 44.9% Model Claude Fable 5 Listed as Fable 5 Conditions Max · Claude Code Interval 41.0–48.7% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,624High WebDev Arena: 1,624 Model Claude Fable 5 Listed as Claude Fable 5 (High) Conditions High Interval 1618–1630 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$1.87 Estimate at list prices Estimate $1.87 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $10, output $50, cache read $1, cache write $12.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Sonnet 4.516. Anthropic	1,012	Pi43%	147.6 Epoch Capabilities Index: 147.6 Model Claude Sonnet 4.5 Listed as Claude Sonnet 4.5 Conditions Not reported Interval 146.2–149.2 Unit ECI Model released Sep 29, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,394High WebDev Arena: 1,394 Model Claude Sonnet 4.5 Listed as Claude Sonnet 4.5 (High, 32K, 20250929) Conditions High Interval 1386–1401 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.56 Estimate at list prices Estimate $0.56 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $3, output $15, cache read $0.30, cache write $3.75 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Codex Auto Review17. OpenAI	975	Codex CLI94%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	No list price on models.dev
GPT-5.2 Codex18. OpenAI	967	Codex CLI92%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	1,339 WebDev Arena: 1,339 Model GPT-5.2 Codex Listed as GPT-5.2-Codex Conditions Not reported Interval 1330–1348 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.35 Estimate at list prices Estimate $0.35 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenRouter: input $1.75, output $14, cache read $0.17, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Gemini 3 Flash Preview19. Google	955	Pi36%	152.3 Epoch Capabilities Index: 152.3 Model Gemini 3 Flash Preview Listed as Gemini 3 Flash Conditions Not reported Interval 150.8–154.0 Unit ECI Model released Dec 17, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	$0.09 Estimate at list prices Estimate $0.09 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Google: input $0.50, output $3, cache read $0.05, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GPT-6 Astra20. OpenAI	937	Codex CLI83%	169.0 Epoch Capabilities Index: 169.0 Model GPT-6 Astra Listed as GPT-6 Astra Conditions Not reported Interval 166.1–172.3 Unit ECI Model released Sep 3, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	60.3%Max · Codex Terminal-Bench 4.1: 60.3% Model GPT-6 Astra Listed as GPT-6 Astra Conditions Max · Codex Interval 57.5–63.1% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Best of 5 reported results for this model. Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,774Max WebDev Arena: 1,774 Model GPT-6 Astra Listed as GPT-6 Astra (Max) Conditions Max Interval 1764–1784 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$1.87 Estimate at list prices Estimate $1.87 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $10, output $50, cache read $1, cache write $12.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Opus 4.521. Anthropic	877	Claude Code62%	151.9 Epoch Capabilities Index: 151.9 Model Claude Opus 4.5 Listed as Claude Opus 4.5 Conditions Not reported Interval 150.3–153.6 Unit ECI Model released Nov 24, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,494High WebDev Arena: 1,494 Model Claude Opus 4.5 Listed as Claude Opus 4.5 (High, 32K, 20251101) Conditions High Interval 1485–1502 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.94 Estimate at list prices Estimate $0.94 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Opus 5.522. Anthropic	860	Claude Code83%	168.5 Epoch Capabilities Index: 168.5 Model Claude Opus 5.5 Listed as Claude Opus 5.5 Conditions Not reported Interval 165.6–171.8 Unit ECI Model released Sep 22, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	65.8%Max · Claude Code Terminal-Bench 4.1: 65.8% Model Claude Opus 5.5 Listed as Opus 5.5 Conditions Max · Claude Code Interval 62.6–68.9% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,807Max WebDev Arena: 1,807 Model Claude Opus 5.5 Listed as Claude Opus 5.5 (Max) Conditions Max Interval 1793–1821 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.56 Estimate at list prices Estimate $0.56 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $4, output $20, cache read $0.20, cache write $5 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-6.1 Sol23. OpenAI	777	Codex CLI79%	168.5 Epoch Capabilities Index: 168.5 Model GPT-6.1 Sol Listed as GPT-6.1 Sol Conditions Not reported Interval 165.5–171.8 Unit ECI Model released Sep 29, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	60.9%Max · Codex Terminal-Bench 4.1: 60.9% Model GPT-6.1 Sol Listed as GPT-6.1 Sol Conditions Max · Codex Interval 57.9–63.9% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,753Max WebDev Arena: 1,753 Model GPT-6.1 Sol Listed as GPT-6.1 Sol (Max) Conditions Max Interval 1740–1766 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.28 Estimate at list prices Estimate $0.28 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $2, output $10, cache read $0.10, cache write $2.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Big Pickle24. Lab unknown	742	OpenCode90%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	No list price on models.dev
GPT-5.225. OpenAI	741	Codex CLI84%	154.5 Epoch Capabilities Index: 154.5 Model GPT-5.2 Listed as GPT-5.2 Conditions Not reported Interval 152.8–156.2 Unit ECI Model released Dec 11, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,417 WebDev Arena: 1,417 Model GPT-5.2 Listed as GPT-5.2 Conditions Not reported Interval 1394–1439 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.35 Estimate at list prices Estimate $0.35 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $1.75, output $14, cache read $0.17, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Gemini 3 Pro Preview26. Google	676	Gemini CLI57%	153.0 Epoch Capabilities Index: 153.0 Model Gemini 3 Pro Preview Listed as Gemini 3 Pro Conditions Not reported Interval 151.4–154.7 Unit ECI Model released Nov 18, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	No list price on models.dev
Claude Fable 5.127. Anthropic	671	Claude Code76%	167.0 Epoch Capabilities Index: 167.0 Model Claude Fable 5.1 Listed as Claude Fable 5.1 Conditions Not reported Interval 164.2–170.1 Unit ECI Model released Sep 1, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	57.6%XHigh · Claude Code Terminal-Bench 4.1: 57.6% Model Claude Fable 5.1 Listed as Fable 5.1 Conditions XHigh · Claude Code Interval 54.3–60.9% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Best of 5 reported results for this model. Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,743Max WebDev Arena: 1,743 Model Claude Fable 5.1 Listed as Claude Fable 5.1 (Max) Conditions Max Interval 1734–1752 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$1.17 Estimate at list prices Estimate $1.17 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $10, output $50, cache read $0.25, cache write $12.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Sonnet 5.528. Anthropic	669	Claude Code79%	166.6 Epoch Capabilities Index: 166.6 Model Claude Sonnet 5.5 Listed as Claude Sonnet 5.5 Conditions Not reported Interval 163.8–169.6 Unit ECI Model released Sep 28, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	63.9%Max · Claude Code Terminal-Bench 4.1: 63.9% Model Claude Sonnet 5.5 Listed as Sonnet 5.5 Conditions Max · Claude Code Interval 60.9–66.9% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,768XHigh WebDev Arena: 1,768 Model Claude Sonnet 5.5 Listed as Claude Sonnet 5.5 (xHigh) Conditions XHigh Interval 1756–1781 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.28 Estimate at list prices Estimate $0.28 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $2, output $10, cache read $0.10, cache write $2.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.3 Codex Spark29. OpenAI	647	Codex CLI86%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	$0.35 Estimate at list prices Estimate $0.35 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $1.75, output $14, cache read $0.17, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GPT-6 Luna30. OpenAI	642	Codex CLI85%	156.4 Epoch Capabilities Index: 156.4 Model GPT-6 Luna Listed as GPT-6 Luna Conditions Not reported Interval 154.4–158.5 Unit ECI Model released Sep 22, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	15.2%Max · Codex Terminal-Bench 4.1: 15.2% Model GPT-6 Luna Listed as GPT-6 Luna Conditions Max · Codex Interval 12.5–17.8% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,578Max WebDev Arena: 1,578 Model GPT-6 Luna Listed as GPT-6 Luna (Max) Conditions Max Interval 1572–1584 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.02 Estimate at list prices Estimate $0.02 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $0.10, output $0.50, cache read $0.01, cache write $0.13 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-6 Sol31. OpenAI	639	Codex CLI83%	163.8 Epoch Capabilities Index: 163.8 Model GPT-6 Sol Listed as GPT-6 Sol Conditions Not reported Interval 161.3–166.4 Unit ECI Model released Sep 22, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	50.3%Max · Codex Terminal-Bench 4.1: 50.3% Model GPT-6 Sol Listed as GPT-6 Sol Conditions Max · Codex Interval 47.1–53.5% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,683Max WebDev Arena: 1,683 Model GPT-6 Sol Listed as GPT-6 Sol (Max) Conditions Max Interval 1673–1693 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.37 Estimate at list prices Estimate $0.37 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $2, output $10, cache read $0.20, cache write $2.50 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
DeepSeek V4 Flash32. DeepSeek	622	Pi31%	146.8 Epoch Capabilities Index: 146.8 Model DeepSeek V4 Flash Listed as DeepSeek-V4-Flash Conditions Not reported Interval 143.6–150.1 Unit ECI Model released Apr 24, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,582High WebDev Arena: 1,582 Model DeepSeek V4 Flash Listed as DeepSeek-V4-Flash (High) Conditions High Interval 1573–1591 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.19 Estimate at list prices Estimate $0.19 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Azure: input $0.19, output $0.51, cache read —, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-read price is listed, so cache reads are priced as input. No cache-write price is listed, so cache writes are priced as input.
DeepSeek V4 Pro33. DeepSeek	615	OpenCode29%	150.2 Epoch Capabilities Index: 150.2 Model DeepSeek V4 Pro Listed as DeepSeek-V4-Pro Conditions Not reported Interval 148.4–152.0 Unit ECI Model released Apr 24, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,582High WebDev Arena: 1,582 Model DeepSeek V4 Pro Listed as DeepSeek-V4-Pro (High, 20260813) Conditions High Interval 1573–1592 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$1.75 Estimate at list prices Estimate $1.75 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Azure: input $1.74, output $3.48, cache read —, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-read price is listed, so cache reads are priced as input. No cache-write price is listed, so cache writes are priced as input.
Gemini 3.1 Pro Preview34. Google	611	Pi78%	156.2 Epoch Capabilities Index: 156.2 Model Gemini 3.1 Pro Preview Listed as Gemini 3.1 Pro Conditions Not reported Interval 154.4–158.1 Unit ECI Model released Feb 19, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,447 WebDev Arena: 1,447 Model Gemini 3.1 Pro Preview Listed as Gemini 3.1 Pro Preview Conditions Not reported Interval 1442–1452 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.38 Estimate at list prices Estimate $0.38 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Google: input $2, output $12, cache read $0.20, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Gemini 2.5 Pro35. Google	607	Gemini CLI89%	144.9 Epoch Capabilities Index: 144.9 Model Gemini 2.5 Pro Listed as Gemini 2.5 Pro (Jun 2025) Conditions Not reported Interval 143.4–146.3 Unit ECI Model released Jun 5, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,226 WebDev Arena: 1,226 Model Gemini 2.5 Pro Listed as Gemini 2.5 Pro Conditions Not reported Interval 1210–1242 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.25 Estimate at list prices Estimate $0.25 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Google: input $1.25, output $10, cache read $0.13, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Gemini 2.5 Flash36. Google	590	Gemini CLI77%	140.8 Epoch Capabilities Index: 140.8 Model Gemini 2.5 Flash Listed as Gemini 2.5 Flash (Jun 2025) Conditions Not reported Interval 138.4–143.0 Unit ECI Model released Jun 17, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	$0.06 Estimate at list prices Estimate $0.06 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Google: input $0.30, output $2.50, cache read $0.03, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GLM-5.237. Z.ai	550	OpenCode23%	153.0 Epoch Capabilities Index: 153.0 Model GLM-5.2 Listed as GLM-5.2 Conditions Not reported Interval 151.3–154.8 Unit ECI Model released Jun 16, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,604Max WebDev Arena: 1,604 Model GLM-5.2 Listed as GLM-5.2 (Max) Conditions Max Interval 1598–1610 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.33 Estimate at list prices Estimate $0.33 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Z.AI: input $1.40, output $4.40, cache read $0.26, cache write $0 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Claude Haiku 5.538. Anthropic	528	Claude Code87%	160.4 Epoch Capabilities Index: 160.4 Model Claude Haiku 5.5 Listed as Claude Haiku 5.5 Conditions Not reported Interval 158.3–162.9 Unit ECI Model released Oct 7, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,586High WebDev Arena: 1,586 Model Claude Haiku 5.5 Listed as Claude Haiku 5.5 (High) Conditions High Interval 1575–1597 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.02 Estimate at list prices Estimate $0.02 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Anthropic: input $0.10, output $0.50, cache read $0.01, cache write $0.13 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5 Codex39. OpenAI	497	Codex CLI99%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	$0.26 Estimate at list prices Estimate $0.26 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Vercel AI Gateway: input $1.25, output $10, cache read $0.13, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GPT-5.1 Codex Max40. OpenAI	495	Codex CLI98%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	$0.25 Estimate at list prices Estimate $0.25 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Azure: input $1.25, output $10, cache read $0.13, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GPT-5.1 Codex Mini41. OpenAI	464	Codex CLI82%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	1,244 WebDev Arena: 1,244 Model GPT-5.1 Codex Mini Listed as GPT-5.1-Codex-Mini Conditions Not reported Interval 1222–1266 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.05 Estimate at list prices Estimate $0.05 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenRouter: input $0.25, output $2, cache read $0.03, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
MiniMax M2.5 Free42. MiniMax	442	OpenCode92%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	No list price on models.dev
GLM-5.143. Z.ai	430	OpenCode48%	150.7 Epoch Capabilities Index: 150.7 Model GLM-5.1 Listed as GLM-5.1 Conditions Not reported Interval 148.6–152.9 Unit ECI Model released Apr 7, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,508 WebDev Arena: 1,508 Model GLM-5.1 Listed as GLM-5.1 Conditions Not reported Interval 1501–1515 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.33 Estimate at list prices Estimate $0.33 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Z.AI: input $1.40, output $4.40, cache read $0.26, cache write $0 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GLM-5.3 Flash44. Z.ai	404	Oh My Pi30%	153.8 Epoch Capabilities Index: 153.8 Model GLM-5.3 Flash Listed as GLM-5.3-Flash Conditions Not reported Interval 152.0–155.8 Unit ECI Model released Aug 20, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	35.4%No reasoning · Claude Code Terminal-Bench 4.1: 35.4% Model GLM-5.3 Flash Listed as GLM-5.3-Flash Conditions No reasoning · Claude Code Interval 31.9–39.0% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,607 WebDev Arena: 1,607 Model GLM-5.3 Flash Listed as GLM-5.3-Flash Conditions Not reported Interval 1600–1613 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.04 Estimate at list prices Estimate $0.04 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Z.AI: input $0.15, output $0.50, cache read $0.03, cache write $0 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
Kimi K345. Moonshot AI	370	Claude Code28%	158.6 Epoch Capabilities Index: 158.6 Model Kimi K3 Listed as Kimi K3 Conditions Not reported Interval 156.6–160.8 Unit ECI Model released Jul 16, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,655Max WebDev Arena: 1,655 Model Kimi K3 Listed as Kimi K3 (Max) Conditions Max Interval 1649–1661 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.55 Estimate at list prices Estimate $0.55 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Moonshot AI: input $3, output $15, cache read $0.30, cache write $3 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
DeepSeek V4 Flash Free46. DeepSeek	364	OpenCode73%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	No list price on models.dev
GLM-5.347. Z.ai	359	Claude Code28%	157.1 Epoch Capabilities Index: 157.1 Model GLM-5.3 Listed as GLM-5.3 Conditions Not reported Interval 154.9–159.5 Unit ECI Model released Aug 14, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	41.2%Max · Claude Code Terminal-Bench 4.1: 41.2% Model GLM-5.3 Listed as GLM-5.3 Conditions Max · Claude Code Interval 37.9–44.5% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,621Max WebDev Arena: 1,621 Model GLM-5.3 Listed as GLM-5.3 (Max) Conditions Max Interval 1614–1629 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.33 Estimate at list prices Estimate $0.33 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Z.AI: input $1.40, output $4.40, cache read $0.26, cache write $0 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GLM-4.748. Z.ai	353	OpenCode48%	143.3 Epoch Capabilities Index: 143.3 Model GLM-4.7 Listed as GLM-4.7 Conditions Not reported Interval 141.3–145.3 Unit ECI Model released Dec 22, 2025 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,435 WebDev Arena: 1,435 Model GLM-4.7 Listed as GLM-4.7 Conditions Not reported Interval 1423–1447 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.14 Estimate at list prices Estimate $0.14 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Z.AI: input $0.60, output $2.20, cache read $0.11, cache write $0 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
DeepSeek V4.1 Flash49. DeepSeek	340	OpenCode27%	156.7 Epoch Capabilities Index: 156.7 Model DeepSeek V4.1 Flash Listed as DeepSeek V4.1 Flash Conditions Not reported Interval 154.8–158.8 Unit ECI Model released Sep 9, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,664Max WebDev Arena: 1,664 Model DeepSeek V4.1 Flash Listed as DeepSeek-V4.1-Flash-20261017 (Max) Conditions Max Interval 1650–1679 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.03 Estimate at list prices Estimate $0.03 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Azure: input $0.30, output $1.20, cache read $0.006, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Kimi K2.650. Moonshot AI	335	OpenCode72%	151.6 Epoch Capabilities Index: 151.6 Model Kimi K2.6 Listed as Kimi K2.6 Conditions Not reported Interval 149.8–153.5 Unit ECI Model released Apr 20, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,509 WebDev Arena: 1,509 Model Kimi K2.6 Listed as Kimi K2.6 Conditions Not reported Interval 1502–1517 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.23 Estimate at list prices Estimate $0.23 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Moonshot AI: input $0.95, output $4, cache read $0.16, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
MiniMax M2.751. MiniMax	328	Pi53%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	1,399 WebDev Arena: 1,399 Model MiniMax M2.7 Listed as MiniMax-M2.7 Conditions Not reported Interval 1393–1404 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.08 Estimate at list prices Estimate $0.08 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price MiniMax (minimax.io): input $0.30, output $1.20, cache read $0.06, cache write $0.38 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GLM-4.7 Free52. Z.ai	328	OpenCode98%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	Not evaluated on WebDev Arena	No list price on models.dev
MiniMax M353. MiniMax	324	OpenCode29%	147.7 Epoch Capabilities Index: 147.7 Model MiniMax M3 Listed as MiniMax-M3 Conditions Not reported Interval 145.6–150.0 Unit ECI Model released Jun 1, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,482 WebDev Arena: 1,482 Model MiniMax M3 Listed as MiniMax-M3 Conditions Not reported Interval 1477–1488 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.08 Estimate at list prices Estimate $0.08 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price MiniMax (minimax.io): input $0.30, output $1.20, cache read $0.06, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GLM-554. Z.ai	323	OpenCode59%	146.5 Epoch Capabilities Index: 146.5 Model GLM-5 Listed as GLM-5 Conditions Not reported Interval 144.8–148.2 Unit ECI Model released Feb 11, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,435 WebDev Arena: 1,435 Model GLM-5 Listed as GLM-5 Conditions Not reported Interval 1427–1444 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.25 Estimate at list prices Estimate $0.25 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Z.AI: input $1, output $3.20, cache read $0.20, cache write $0 per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
GPT-5.1 Codex55. OpenAI	312	Codex CLI97%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	1,337 WebDev Arena: 1,337 Model GPT-5.1 Codex Listed as GPT-5.1-Codex Conditions Not reported Interval 1326–1349 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.26 Estimate at list prices Estimate $0.26 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenRouter: input $1.25, output $10, cache read $0.13, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Grok 4.556. xAI	305	Grok Build50%	155.0 Epoch Capabilities Index: 155.0 Model Grok 4.5 Listed as Grok 4.5 Conditions Not reported Interval 153.2–156.9 Unit ECI Model released Jul 8, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	12.4%High · Grok Build Terminal-Bench 4.1: 12.4% Model Grok 4.5 Listed as Grok 4.5 Conditions High · Grok Build Interval 9.8–15.0% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,553 WebDev Arena: 1,553 Model Grok 4.5 Listed as Grok 4.5 Conditions Not reported Interval 1546–1560 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.44 Estimate at list prices Estimate $0.44 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price xAI: input $2, output $6, cache read $0.30, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Grok 4.657. xAI	302	Grok Build62%	157.8 Epoch Capabilities Index: 157.8 Model Grok 4.6 Listed as Grok 4.6 Conditions Not reported Interval 155.8–159.9 Unit ECI Model released Aug 12, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	19.7%High · Grok Build Terminal-Bench 4.1: 19.7% Model Grok 4.6 Listed as Grok 4.6 Conditions High · Grok Build Interval 16.6–22.8% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,617High WebDev Arena: 1,617 Model Grok 4.6 Listed as Grok 4.6 (High) Conditions High Interval 1609–1624 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.62 Estimate at list prices Estimate $0.62 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price xAI: input $2, output $6, cache read $0.50, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
GPT-558. OpenAI	297	Claude Code79%	Not evaluated on Epoch Capabilities Index	Not evaluated on Terminal-Bench 4.1	1,418Medium WebDev Arena: 1,418 Model GPT-5 Listed as GPT-5 (Medium) Conditions Medium Interval 1402–1433 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.25 Estimate at list prices Estimate $0.25 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price OpenAI: input $1.25, output $10, cache read $0.13, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.
Kimi K2.559. Moonshot AI	295	OpenCode59%	148.8 Epoch Capabilities Index: 148.8 Model Kimi K2.5 Listed as Kimi K2.5 Conditions Not reported Interval 147.2–150.3 Unit ECI Model released Jan 27, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	Not evaluated on Terminal-Bench 4.1	1,437 WebDev Arena: 1,437 Model Kimi K2.5 Listed as Kimi K2.5 (Thinking) Conditions Not reported Interval 1431–1442 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Best of 2 reported results for this model. Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.61 Estimate at list prices Estimate $0.61 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Amazon Bedrock: input $0.60, output $3, cache read —, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-read price is listed, so cache reads are priced as input. No cache-write price is listed, so cache writes are priced as input.
Gemini 3.8 Flash60. Google	286	Antigravity CLI51%	158.8 Epoch Capabilities Index: 158.8 Model Gemini 3.8 Flash Listed as Gemini 3.8 Flash Conditions Not reported Interval 156.8–160.9 Unit ECI Model released Sep 2, 2026 Retrieved Nov 12, 2026 Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)	18.2%High · mini-SWE-agent Terminal-Bench 4.1: 18.2% Model Gemini 3.8 Flash Listed as Gemini 3.8 Flash Conditions High · mini-SWE-agent Interval 14.8–21.5% Unit % resolved Published Nov 3, 2026 Retrieved Nov 12, 2026 Source · Terminal-Bench 4.1 official leaderboard, tbench.ai	1,585High WebDev Arena: 1,585 Model Gemini 3.8 Flash Listed as Gemini 3.8 Flash (High) Conditions High Interval 1577–1592 Unit Arena score Published Oct 19, 2026 Retrieved Oct 21, 2026 Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0	$0.14 Estimate at list prices Estimate $0.14 per 1M tokens Mix the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1% List price Google: input $0.75, output $3.75, cache read $0.07, cache write — per 1M The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid. No cache-write price is listed, so cache writes are priced as input.

1
Claude Haiku 4.5Anthropic
Active users
1,807
Top agent
Claude Code91%
ECI
142.8
Epoch Capabilities Index: 142.8
Model
Claude Haiku 4.5
Listed as
Claude Haiku 4.5
Conditions
Not reported
Interval
141.1–144.5
Unit
ECI
Model released
Oct 15, 2025
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,330
WebDev Arena: 1,330
Model
Claude Haiku 4.5
Listed as
Claude Haiku 4.5 (20251001)
Conditions
Not reported
Interval
1325–1335
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.19
Estimate at list prices
Estimate
$0.19 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $1, output $5, cache read $0.10, cache write $1.25 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
2
GPT-5.5OpenAI
Active users
1,586
Top agent
Codex CLI89%
ECI
160.1
Epoch Capabilities Index: 160.1
Model
GPT-5.5
Listed as
GPT-5.5
Conditions
Not reported
Interval
158.1–162.4
Unit
ECI
Model released
Apr 23, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,513XHigh · Codex
WebDev Arena: 1,513
Model
GPT-5.5
Listed as
GPT-5.5 (xHigh, Codex Harness)
Conditions
XHigh · Codex
Interval
1507–1519
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 3 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.94
Estimate at list prices
Estimate
$0.94 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $5, output $30, cache read $0.50, cache write — per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
No cache-write price is listed, so cache writes are priced as input.
3
Claude Sonnet 4.6Anthropic
Active users
1,566
Top agent
Claude Code71%
ECI
152.9
Epoch Capabilities Index: 152.9
Model
Claude Sonnet 4.6
Listed as
Claude Sonnet 4.6
Conditions
Not reported
Interval
151.2–154.7
Unit
ECI
Model released
Feb 17, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,523
WebDev Arena: 1,523
Model
Claude Sonnet 4.6
Listed as
Claude Sonnet 4.6
Conditions
Not reported
Interval
1518–1528
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.56
Estimate at list prices
Estimate
$0.56 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $3, output $15, cache read $0.30, cache write $3.75 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
4
GPT-5.4OpenAI
Active users
1,529
Top agent
Codex CLI88%
ECI
157.8
Epoch Capabilities Index: 157.8
Model
GPT-5.4
Listed as
GPT-5.4
Conditions
Not reported
Interval
155.9–159.9
Unit
ECI
Model released
Mar 5, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,464High · Codex
WebDev Arena: 1,464
Model
GPT-5.4
Listed as
GPT-5.4 (High, Codex Harness)
Conditions
High · Codex
Interval
1445–1483
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 3 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.47
Estimate at list prices
Estimate
$0.47 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $2.50, output $15, cache read $0.25, cache write — per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
No cache-write price is listed, so cache writes are priced as input.
5
Claude Opus 4.6Anthropic
Active users
1,414
Top agent
Claude Code78%
ECI
156.3
Epoch Capabilities Index: 156.3
Model
Claude Opus 4.6
Listed as
Claude Opus 4.6
Conditions
Not reported
Interval
154.6–158.3
Unit
ECI
Model released
Feb 5, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,547High
WebDev Arena: 1,547
Model
Claude Opus 4.6
Listed as
Claude Opus 4.6 (High)
Conditions
High
Interval
1541–1552
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 2 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.94
Estimate at list prices
Estimate
$0.94 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
6
GPT-5.3 CodexOpenAI
Active users
1,335
Top agent
Codex CLI85%
ECI
Not evaluated on Epoch Capabilities Index
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,407Codex
WebDev Arena: 1,407
Model
GPT-5.3 Codex
Listed as
GPT-5.3-Codex (Codex Harness)
Conditions
Codex
Interval
1393–1420
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 2 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.35
Estimate at list prices
Estimate
$0.35 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $1.75, output $14, cache read $0.17, cache write — per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
No cache-write price is listed, so cache writes are priced as input.
7
GPT-5.6 SolOpenAI
Active users
1,262
Top agent
Codex CLI85%
ECI
163.0
Epoch Capabilities Index: 163.0
Model
GPT-5.6 Sol
Listed as
GPT-5.6 Sol
Conditions
Not reported
Interval
160.7–165.5
Unit
ECI
Model released
Jul 9, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
38.5%Max · Codex
Terminal-Bench 4.1: 38.5%
Model
GPT-5.6 Sol
Listed as
GPT-5.6 Sol
Conditions
Max · Codex
Interval
34.6–42.3%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,616XHigh · Codex
WebDev Arena: 1,616
Model
GPT-5.6 Sol
Listed as
GPT-5.6 Sol (xHigh, Codex Harness)
Conditions
XHigh · Codex
Interval
1610–1622
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.75
Estimate at list prices
Estimate
$0.75 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $4, output $20, cache read $0.40, cache write $5 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
8
Claude Opus 4.8Anthropic
Active users
1,258
Top agent
Claude Code87%
ECI
159.9
Epoch Capabilities Index: 159.9
Model
Claude Opus 4.8
Listed as
Claude Opus 4.8
Conditions
Not reported
Interval
157.9–162.2
Unit
ECI
Model released
May 28, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
23.6%Max · Claude Code
Terminal-Bench 4.1: 23.6%
Model
Claude Opus 4.8
Listed as
Opus 4.8
Conditions
Max · Claude Code
Interval
20.1–27.2%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,556High
WebDev Arena: 1,556
Model
Claude Opus 4.8
Listed as
Claude Opus 4.8 (High)
Conditions
High
Interval
1550–1562
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 2 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.94
Estimate at list prices
Estimate
$0.94 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
9
GPT-5.6 TerraOpenAI
Active users
1,106
Top agent
Codex CLI71%
ECI
160.7
Epoch Capabilities Index: 160.7
Model
GPT-5.6 Terra
Listed as
GPT-5.6 Terra
Conditions
Not reported
Interval
158.5–163.3
Unit
ECI
Model released
Jul 9, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
21.5%Max · Codex
Terminal-Bench 4.1: 21.5%
Model
GPT-5.6 Terra
Listed as
GPT-5.6 Terra
Conditions
Max · Codex
Interval
18.3–24.8%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,522XHigh · Codex
WebDev Arena: 1,522
Model
GPT-5.6 Terra
Listed as
GPT-5.6 Terra (xHigh, Codex Harness)
Conditions
XHigh · Codex
Interval
1515–1528
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.39
Estimate at list prices
Estimate
$0.39 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $2, output $12, cache read $0.20, cache write $2.50 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
10
GPT-5.6 LunaOpenAI
Active users
1,079
Top agent
Codex CLI74%
ECI
156.7
Epoch Capabilities Index: 156.7
Model
GPT-5.6 Luna
Listed as
GPT-5.6 Luna
Conditions
Not reported
Interval
154.8–158.7
Unit
ECI
Model released
Jul 9, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
17.6%Max · Codex
Terminal-Bench 4.1: 17.6%
Model
GPT-5.6 Luna
Listed as
GPT-5.6 Luna
Conditions
Max · Codex
Interval
14.7–20.5%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,518XHigh · Codex
WebDev Arena: 1,518
Model
GPT-5.6 Luna
Listed as
GPT-5.6 Luna (xHigh, Codex Harness)
Conditions
XHigh · Codex
Interval
1512–1525
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.04
Estimate at list prices
Estimate
$0.04 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $0.20, output $1.20, cache read $0.02, cache write $0.25 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
11
Claude Sonnet 5Anthropic
Active users
1,056
Top agent
Claude Code89%
ECI
156.9
Epoch Capabilities Index: 156.9
Model
Claude Sonnet 5
Listed as
Claude Sonnet 5
Conditions
Not reported
Interval
154.8–159.2
Unit
ECI
Model released
Jun 30, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
12.1%Max · Claude Code
Terminal-Bench 4.1: 12.1%
Model
Claude Sonnet 5
Listed as
Sonnet 5
Conditions
Max · Claude Code
Interval
9.1–15.1%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,540High
WebDev Arena: 1,540
Model
Claude Sonnet 5
Listed as
Claude Sonnet 5 (High)
Conditions
High
Interval
1534–1546
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.37
Estimate at list prices
Estimate
$0.37 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $2, output $10, cache read $0.20, cache write $2.50 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
12
Claude Opus 5Anthropic
Active users
1,050
Top agent
Claude Code88%
ECI
164.7
Epoch Capabilities Index: 164.7
Model
Claude Opus 5
Listed as
Claude Opus 5
Conditions
Not reported
Interval
162.1–167.5
Unit
ECI
Model released
Jul 24, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
53.3%XHigh · Claude Code
Terminal-Bench 4.1: 53.3%
Model
Claude Opus 5
Listed as
Opus 5
Conditions
XHigh · Claude Code
Interval
50.2–56.4%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Best of 5 reported results for this model.
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,689Max
WebDev Arena: 1,689
Model
Claude Opus 5
Listed as
Claude Opus 5 (Max)
Conditions
Max
Interval
1683–1695
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 2 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.94
Estimate at list prices
Estimate
$0.94 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
13
Claude Opus 4.7Anthropic
Active users
1,043
Top agent
Claude Code87%
ECI
157.4
Epoch Capabilities Index: 157.4
Model
Claude Opus 4.7
Listed as
Claude Opus 4.7
Conditions
Not reported
Interval
155.5–159.6
Unit
ECI
Model released
Apr 16, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,558High
WebDev Arena: 1,558
Model
Claude Opus 4.7
Listed as
Claude Opus 4.7 (High)
Conditions
High
Interval
1552–1564
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 2 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.94
Estimate at list prices
Estimate
$0.94 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $5, output $25, cache read $0.50, cache write $6.25 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
14
GPT-5.4 MiniOpenAI
Active users
1,041
Top agent
Codex CLI74%
ECI
149.3
Epoch Capabilities Index: 149.3
Model
GPT-5.4 Mini
Listed as
GPT-5.4 Mini
Conditions
Not reported
Interval
147.7–151.1
Unit
ECI
Model released
Mar 17, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,397High
WebDev Arena: 1,397
Model
GPT-5.4 Mini
Listed as
GPT-5.4 mini (High)
Conditions
High
Interval
1391–1404
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.14
Estimate at list prices
Estimate
$0.14 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $0.75, output $4.50, cache read $0.07, cache write — per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
No cache-write price is listed, so cache writes are priced as input.
15
Claude Fable 5Anthropic
Active users
1,020
Top agent
Claude Code87%
ECI
164.1
Epoch Capabilities Index: 164.1
Model
Claude Fable 5
Listed as
Claude Fable 5
Conditions
Not reported
Interval
161.7–166.8
Unit
ECI
Model released
Jun 9, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
44.9%Max · Claude Code
Terminal-Bench 4.1: 44.9%
Model
Claude Fable 5
Listed as
Fable 5
Conditions
Max · Claude Code
Interval
41.0–48.7%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,624High
WebDev Arena: 1,624
Model
Claude Fable 5
Listed as
Claude Fable 5 (High)
Conditions
High
Interval
1618–1630
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$1.87
Estimate at list prices
Estimate
$1.87 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $10, output $50, cache read $1, cache write $12.50 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
16
Claude Sonnet 4.5Anthropic
Active users
1,012
Top agent
Pi43%
ECI
147.6
Epoch Capabilities Index: 147.6
Model
Claude Sonnet 4.5
Listed as
Claude Sonnet 4.5
Conditions
Not reported
Interval
146.2–149.2
Unit
ECI
Model released
Sep 29, 2025
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,394High
WebDev Arena: 1,394
Model
Claude Sonnet 4.5
Listed as
Claude Sonnet 4.5 (High, 32K, 20250929)
Conditions
High
Interval
1386–1401
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Best of 2 reported results for this model.
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.56
Estimate at list prices
Estimate
$0.56 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Anthropic: input $3, output $15, cache read $0.30, cache write $3.75 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
17
Codex Auto ReviewOpenAI
Active users
975
Top agent
Codex CLI94%
ECI
Not evaluated on Epoch Capabilities Index
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
Not evaluated on WebDev Arena
$/1M at agent mix
No list price on models.dev
18
GPT-5.2 CodexOpenAI
Active users
967
Top agent
Codex CLI92%
ECI
Not evaluated on Epoch Capabilities Index
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
1,339
WebDev Arena: 1,339
Model
GPT-5.2 Codex
Listed as
GPT-5.2-Codex
Conditions
Not reported
Interval
1330–1348
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$0.35
Estimate at list prices
Estimate
$0.35 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenRouter: input $1.75, output $14, cache read $0.17, cache write — per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
No cache-write price is listed, so cache writes are priced as input.
19
Gemini 3 Flash PreviewGoogle
Active users
955
Top agent
Pi36%
ECI
152.3
Epoch Capabilities Index: 152.3
Model
Gemini 3 Flash Preview
Listed as
Gemini 3 Flash
Conditions
Not reported
Interval
150.8–154.0
Unit
ECI
Model released
Dec 17, 2025
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
Not evaluated on Terminal-Bench 4.1
WebDev Arena
Not evaluated on WebDev Arena
$/1M at agent mix
$0.09
Estimate at list prices
Estimate
$0.09 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
Google: input $0.50, output $3, cache read $0.05, cache write — per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.
No cache-write price is listed, so cache writes are priced as input.
20
GPT-6 AstraOpenAI
Active users
937
Top agent
Codex CLI83%
ECI
169.0
Epoch Capabilities Index: 169.0
Model
GPT-6 Astra
Listed as
GPT-6 Astra
Conditions
Not reported
Interval
166.1–172.3
Unit
ECI
Model released
Sep 3, 2026
Retrieved
Nov 12, 2026
Source · Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Terminal-Bench 4
60.3%Max · Codex
Terminal-Bench 4.1: 60.3%
Model
GPT-6 Astra
Listed as
GPT-6 Astra
Conditions
Max · Codex
Interval
57.5–63.1%
Unit
% resolved
Published
Nov 3, 2026
Retrieved
Nov 12, 2026
Best of 5 reported results for this model.
Source · Terminal-Bench 4.1 official leaderboard, tbench.ai
WebDev Arena
1,774Max
WebDev Arena: 1,774
Model
GPT-6 Astra
Listed as
GPT-6 Astra (Max)
Conditions
Max
Interval
1764–1784
Unit
Arena score
Published
Oct 19, 2026
Retrieved
Oct 21, 2026
Source · LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset on Hugging Face), CC BY 4.0
$/1M at agent mix
$1.87
Estimate at list prices
Estimate
$1.87 per 1M tokens
Mix
the community token mix: Cache read 93%, Input 4.6%, Cache write 1.5%, Output 0.5%, Reasoning <0.1%
List price
OpenAI: input $10, output $50, cache read $1, cache write $12.50 per 1M
The list price of each token type, weighted by its share of the mix. Reasoning tokens are priced as output. Discounts, batch prices and long-context tiers are not applied. Not what anyone paid.

Usage counts public Tokscale submissions only; hidden accounts are excluded. Scores are the best result each source reports for the model, with its effort and agent or harness; open a score for its source and date. Prices are list prices from models.dev, and every $/1M figure is an estimate at list prices, not what anyone paid. Usage and scores are shown side by side; one does not explain the other. Sorting by a score puts models that were not evaluated last.

Usage, capability and price

The models in the table that have an Epoch Capabilities Index. These charts put usage and scores side by side; they do not show that one causes the other.

Usage vs capability. Active users in the all time (log scale) against the Epoch Capabilities Index.

Show data

Active users and Epoch Capabilities Index by model
Model	Active users	ECI
Claude Fable 5.2	238	173.4
Gemini 4 Carbon	10	169.8
GPT-6 Astra	937	169.0
Claude Opus 5.5	860	168.5
GPT-6.1 Sol	777	168.5
GPT-6.1 Terra	278	167.7
Claude Fable 5.1	671	167.0
Claude Sonnet 5.5	669	166.6
Gemini 4 Argon	84	166.5
Claude Opus 5	1,050	164.7
Claude Fable 5	1,020	164.1
GPT-6 Sol	639	163.8
GPT-5.6 Sol	1,262	163.0
GPT-5.5 Pro	16	162.7
Muse Wonder 1.0	20	161.7
Kimi K3.1 Pro	42	161.4
GPT-5.6 Terra	1,106	160.7
Claude Haiku 5.5	528	160.4
GPT-5.4 Pro	20	160.2
GPT-5.5	1,586	160.1
Claude Opus 4.8	1,258	159.9
DeepSeek V4.1 Flash 1017	78	159.4
Gemini 3.8 Flash	286	158.8
Kimi K3	370	158.6
Gemini 3.7 Flash	221	158.4
Muse Spark 1.3	47	158.2
GPT-5.4	1,529	157.8
Grok 4.6	302	157.8
Qwen3.8 Max	117	157.6
Claude Opus 4.7	1,043	157.4
GPT-5.2 Pro	42	157.3
GLM-5.3	359	157.1
Claude Sonnet 5	1,056	156.9
DeepSeek V4.1 Flash	340	156.7
GPT-5.6 Luna	1,079	156.7
GPT-6 Luna	642	156.4
Claude Opus 4.6	1,414	156.3
Gemini 3.1 Pro Preview	611	156.2
Grok 4.7	207	156.2
DeepSeek V4 Pro 0813	54	155.9
Gemini 3.5 Flash	140	155.3
Gemini 3.6 Flash	214	155.2
Muse Spark 1.2	27	155.0
Grok 4.5	305	155.0
Muse Spark 1.1	5	154.9
Qwen3.7 Max	118	154.5
GPT-5.2	741	154.5
DeepSeek V4 Flash 0731	154	154.4
GLM-5.3 Flash	404	153.8
GLM-5.2	550	153.0
Gemini 3 Pro Preview	676	153.0
Claude Sonnet 4.6	1,566	152.9
Qwen3.8 27B	68	152.6
Gemini 3 Flash Preview	955	152.3
Grok 4.20 0309 Reasoning	17	152.3
Claude Opus 4.5	877	151.9
Kimi K2.6	335	151.6
GPT-5 Pro	6	151.2
Kimi K2.7 Code	223	151.2
GPT-5.1	161	150.9
GLM-5.1	430	150.7
DeepSeek V4 Pro	615	150.2
GPT-5 2025 08 07	22	150.0
Grok 4.3	65	149.9
Qwen3.6 Max Preview	10	149.8
GPT-5.4 Mini	1,041	149.3
Qwen3.6 Plus	123	148.9
Kimi K2.5	295	148.8
Qwen3.5 397B A17B	22	147.9
Qwen3.7 Plus	93	147.9
MiniMax M3	324	147.7
Claude Sonnet 4.5	1,012	147.6
o3-pro	11	147.5
Qwen3.5 Plus	83	147.3
o3	61	147.1
Grok 4 0709	30	146.9
DeepSeek V4 Flash	622	146.8
Qwen3.6 27B	31	146.7
GLM-5	323	146.5
GPT-5.4 Nano	20	146.4
DeepSeek Chat	58	146.3
GPT-5 Mini	178	146.1
Gemini 3.5 Flash Lite	35	145.6
Kimi K2 Thinking	78	145.4
o4-mini	19	145.3
Gemini 3.1 Flash Lite	63	145.0
Gemini 2.5 Pro	607	144.9
Claude Opus 4.1	75	144.8
Gemma 4 31B IT	37	144.3
Qwen3.6 35B A3B	33	144.0
Qwen3.5 35B A3B	10	144.0
Gemini 2.5 Pro Exp 03 25	7	143.9
Qwen3.6 Flash	11	143.7
GLM-4.7	353	143.3
Qwen3 235B A22B Thinking 2507	5	143.2
Claude Opus 4	28	142.9
Claude Haiku 4.5	1,807	142.8
Qwen3 Max	13	142.6
Claude Sonnet 4	215	141.9
Gemma 4 26B A4B IT	23	141.6
Gemini 2.5 Flash Preview 05 20	8	141.6
Kimi K2 Instruct	17	141.2
Claude 3.7 Sonnet	24	141.0
Gemini 2.5 Pro Preview 05 06	59	140.9
DeepSeek V3.1	10	140.8
Gemini 2.5 Flash	590	140.8
GPT OSS 120B	83	140.2
Qwen3.5 9B	41	139.3
GPT-5 Nano	258	139.1
Qwen3 32B	13	136.9
GPT-4.1 2025 04 14	36	136.2
Qwen3 14B	7	136.1
GPT OSS 20B	55	135.0
GPT-4.1 Mini 2025 04 14	13	134.2
Qwen3 8B	9	132.1
DeepSeek V3	7	131.7
GPT-4o 2024 11 20	10	128.8
Claude 3.5 Haiku	48	124.9
GPT-4o Mini	33	123.9
Claude 3 Haiku	5	117.7

Price vs capability. Estimated USD per 1M tokens at the community token mix in the all time (log scale, list prices; cache reads are 93% of that mix) against ECI. The dashed line is the price frontier: no model is both cheaper and higher on ECI than a model on it.

Show data

Estimated price at the community token mix and Epoch Capabilities Index by model
Model	Est. $/1M	ECI	Frontier
GPT-5 Nano	$0.01	139.1	Yes
DeepSeek V4.1 Flash 1017	$0.02	159.4	Yes
GPT-6 Luna	$0.02	156.4
Claude Haiku 5.5	$0.02	160.4	Yes
DeepSeek V4.1 Flash	$0.03	156.7
DeepSeek V4 Flash 0731	$0.04	154.4
GLM-5.3 Flash	$0.04	153.8
GPT-5.4 Nano	$0.04	146.4
GPT-5.6 Luna	$0.04	156.7
Gemini 3.1 Flash Lite	$0.05	145.0
Qwen3.6 Flash	$0.05	143.7
GPT-5 Mini	$0.05	146.1
Qwen3.8 27B	$0.05	152.6
Gemini 2.5 Flash	$0.06	140.8
Gemini 3.5 Flash Lite	$0.06	145.6
GPT OSS 20B	$0.07	135.0
DeepSeek V4 Pro 0813	$0.07	155.9
MiniMax M3	$0.08	147.7
GPT-4o Mini	$0.08	123.9
Gemini 3 Flash Preview	$0.09	152.3
Qwen3.6 Plus	$0.10	148.9
Qwen3.7 Plus	$0.11	147.9
Gemma 4 26B A4B IT	$0.13	141.6
Gemini 3.8 Flash	$0.14	158.8
Gemini 3.7 Flash	$0.14	158.4
Gemini 3.6 Flash	$0.14	155.2
GPT-6.1 Terra	$0.14	167.7	Yes
Gemma 4 31B IT	$0.14	144.3
GPT-5.4 Mini	$0.14	149.3
GLM-4.7	$0.14	143.3
Qwen3.5 9B	$0.14	139.3
GPT OSS 120B	$0.15	140.2
Qwen3 8B	$0.18	132.1
Claude Haiku 4.5	$0.19	142.8
DeepSeek V4 Flash	$0.19	146.8
Kimi K2.6	$0.23	151.6
Muse Spark 1.3	$0.24	158.2
Muse Spark 1.2	$0.24	155.0
Muse Spark 1.1	$0.24	154.9
Gemini 2.5 Pro	$0.25	144.9
GPT-5.1	$0.25	150.9
Qwen3.6 Max Preview	$0.25	149.8
GLM-5	$0.25	146.5
Kimi K2.7 Code	$0.26	151.2
Qwen3.5 35B A3B	$0.26	144.0
Gemini 4 Argon	$0.27	166.5
Gemini 4 Carbon	$0.27	169.8	Yes
Grok 4.3	$0.28	149.9
Grok 4.20 0309 Reasoning	$0.28	152.3
GPT-6.1 Sol	$0.28	168.5
Claude Sonnet 5.5	$0.28	166.6
Gemini 3.5 Flash	$0.28	155.3
DeepSeek Chat	$0.32	146.3
GLM-5.2	$0.33	153.0
GLM-5.1	$0.33	150.7
GLM-5.3	$0.33	157.1
o4-mini	$0.35	145.3
GPT-5.2	$0.35	154.5
Qwen3 14B	$0.36	136.1
Kimi K3.1 Pro	$0.37	161.4
Claude Sonnet 5	$0.37	156.9
GPT-6 Sol	$0.37	163.8
Gemini 3.1 Pro Preview	$0.38	156.2
GPT-5.6 Terra	$0.39	160.7
Qwen3.6 35B A3B	$0.39	144.0
Qwen3.8 Max	$0.40	157.6
Qwen3.5 Plus	$0.41	147.3
Grok 4.5	$0.44	155.0
Qwen3 235B A22B Thinking 2507	$0.47	143.2
GPT-5.4	$0.47	157.8
Muse Wonder 1.0	$0.55	161.7
Kimi K3	$0.55	158.6
Claude Sonnet 4.6	$0.56	152.9
Claude Sonnet 4.5	$0.56	147.6
Claude Sonnet 4	$0.56	141.9
Claude Opus 5.5	$0.56	168.5
DeepSeek V3.1	$0.59	140.8
Kimi K2 Thinking	$0.61	145.4
Kimi K2.5	$0.61	148.8
Qwen3.6 27B	$0.62	146.7
Qwen3.5 397B A17B	$0.62	147.9
Grok 4.6	$0.62	157.8
Grok 4.7	$0.62	156.2
o3	$0.64	147.1
Qwen3.7 Max	$0.67	154.5
Qwen3 32B	$0.71	136.9
GPT-5.6 Sol	$0.75	163.0
Claude Opus 4.6	$0.94	156.3
Claude Opus 4.8	$0.94	159.9
Claude Opus 5	$0.94	164.7
Claude Opus 4.7	$0.94	157.4
Claude Opus 4.5	$0.94	151.9
Claude Fable 5.2	$0.94	173.4	Yes
GPT-5.5	$0.94	160.1
Claude Fable 5.1	$1.17	167.0
Qwen3 Max	$1.23	142.6
DeepSeek V3	$1.25	131.7
GPT-4o 2024 11 20	$1.38	128.8
DeepSeek V4 Pro	$1.75	150.2
Claude Fable 5	$1.87	164.1
GPT-6 Astra	$1.87	169.0
Claude Opus 4.1	$2.81	144.8
Claude Opus 4	$2.81	142.9
GPT-5 Pro	$15.60	151.2
o3-pro	$20.34	147.5
GPT-5.2 Pro	$21.84	157.3
GPT-5.4 Pro	$30.85	160.2
GPT-5.5 Pro	$30.85	162.7

Sources and attribution

Each score keeps its own source, unit and run conditions. Tokscale does not combine benchmarks into one score, and a missing score means the model was not evaluated, not that it scored zero.

Epoch AIEpoch Capabilities Index, Epoch's own runs and the external results it collects.CC BY 4.0 · Retrieved Nov 12, 2026 · Data through Nov 9, 2026Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from https://epoch.ai/benchmarks.
LMArenaWebDev Arena and Agent Arena leaderboards.CC BY 4.0 · Retrieved Oct 21, 2026 · Data through Oct 19, 2026LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset), Hugging Face.
Terminal-BenchTerminal-Bench 4.1 official leaderboard rows (results only, no task content).Terms not stated · Retrieved Nov 12, 2026 · Data through Nov 11, 2026Terminal-Bench 4.1 leaderboard, tbench.ai. Every score links back to the leaderboard.
models.devSpecs and list prices per provider.MIT · Retrieved Nov 12, 2026models.dev, an open catalog of AI models (github.com/anomalyco/models.dev).
TokscaleActive users, tokens and token mix from public Tokscale submissions. Hidden accounts are excluded.How we count