GPQA Diamond
Hard multiple-choice questions in biology, chemistry and physics, written by PhD-level experts.
- Unit
- % correct
- Results
- 324 results151 Tokscale models matched, 224 in the source
- Last update
- Oct 15, 2026Newest run date
- Retrieved
- Oct 21, 2026Source data through Oct 15, 2026
Conditions. Run by Epoch AI at each reasoning effort.
Attribution. Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
Near its ceiling: the top score is 96.4%, and Epoch AI estimates the reachable maximum at 97%. Differences between top models are small.
Results
One row per model: its best reported result. The rank is the position in this list, by score; the source's own ranking can differ.
Tokscale usage counts accounts that used the model in public Tokscale submissions, Oct 14 – Nov 12, 2026. It is shown next to the score and does not explain it.
| Rank | Model | Score (% correct) | Conditions | Date | Tokscale usage (30 days) | Source link |
|---|---|---|---|---|---|---|
| 1 | Gemini 4 ArgonListed as gemini-4-argon_high | 96.4%Standard error ±1.2 pts | High effort | Oct 13, 2026Run date | 82active users | Source for Gemini 4 Argon (opens in a new tab) |
| 2 | GPT-6 AstraListed as gpt-6-astra_max | 95.8%Standard error ±1.4 pts | Max effort | Aug 30, 2026Run date | 563active users | Source for GPT-6 Astra (opens in a new tab) |
| 3 | Claude Sonnet 5.5Listed as claude-sonnet-5-5_max | 95.6%Standard error ±1.4 pts | Max effort | Sep 29, 2026Run date | 555active users | Source for Claude Sonnet 5.5 (opens in a new tab) |
| 4 | GPT-6.1 SolListed as gpt-6.1-sol_max | 95.4%Standard error ±1.4 pts | Max effort | Sep 29, 2026Run date | 660active users | Source for GPT-6.1 Sol (opens in a new tab) |
| 4 | Gemini 3.8 FlashListed as gemini-3.8-flash_high | 95.4%Standard error ±1.4 pts | High effort | Sep 2, 2026Run date | 66active users | Source for Gemini 3.8 Flash (opens in a new tab) |
| 6 | Gemini 3.7 FlashListed as gemini-3.7-flash_high | 94.8%Standard error ±1.3 pts | High effort | Aug 14, 2026Run date | 4active users | Source for Gemini 3.7 Flash (opens in a new tab) |
| 7 | GPT-5.4 ProListed as gpt-5.4-pro-2026-03-05_xhigh | 94.6%Standard error ±1.6 pts | XHigh effort | Mar 20, 2026Run date | No Tokscale usage in the last 30 days | Source for GPT-5.4 Pro (opens in a new tab) |
| 8 | Gemini 3.1 Pro PreviewListed as gemini-3.1-pro-preview_high | 94.4%Standard error ±1.6 pts | High effort | Aug 6, 2026Run date | 5active users | Source for Gemini 3.1 Pro Preview (opens in a new tab) |
| 9 | GPT-6 SolListed as gpt-6-sol_max | 94.3%Standard error ±1.5 pts | Max effort | Sep 22, 2026Run date | 246active users | Source for GPT-6 Sol (opens in a new tab) |
| 10 | Gemini 3.6 FlashListed as gemini-3.6-flash_high | 94.1%Standard error ±1.4 pts | High effort | Aug 2, 2026Run date | 3active users | Source for Gemini 3.6 Flash (opens in a new tab) |
| 11 | Grok 4.6Listed as grok-4.6_high | 94.0%Standard error ±1.4 pts | High effort | Aug 12, 2026Run date | 41active users | Source for Grok 4.6 (opens in a new tab) |
| 11 | GPT-5.5Listed as gpt-5.5-pre-release_xhigh | 94.0%Standard error ±1.5 pts | XHigh effort | Apr 24, 2026Run date | 30active users | Source for GPT-5.5 (opens in a new tab) |
| 13 | GPT-5.5 ProListed as gpt-5.5-pro-pre-release_xhigh | 93.9%Standard error ±1.6 pts | XHigh effort | Apr 24, 2026Run date | 1active users | Source for GPT-5.5 Pro (opens in a new tab) |
| 14 | Claude Opus 5Listed as claude-opus-5_max | 93.9%Standard error ±1.5 pts | Max effort | Jul 24, 2026Run date | 163active users | Source for Claude Opus 5 (opens in a new tab) |
| 15 | GPT-5.6 SolListed as gpt-5.6-sol_max | 93.5%Standard error ±1.6 pts | Max effort | Jul 9, 2026Run date | 217active users | Source for GPT-5.6 Sol (opens in a new tab) |
| 16 | Grok 4.5Listed as grok-4.5_high | 93.4%Standard error ±1.4 pts | High effort | Jul 8, 2026Run date | 9active users | Source for Grok 4.5 (opens in a new tab) |
| 17 | GPT-5.6 TerraListed as gpt-5.6-terra_max | 93.3%Standard error ±1.5 pts | Max effort | Jul 9, 2026Run date | 120active users | Source for GPT-5.6 Terra (opens in a new tab) |
| 18 | GPT-5.4Listed as gpt-5.4-2026-03-05_xhigh | 93.3%Standard error ±1.8 pts | XHigh effort | Mar 6, 2026Run date | 1active users | Source for GPT-5.4 (opens in a new tab) |
| 19 | Kimi K3Listed as kimi-k3_max | 93.1%Standard error ±1.5 pts | Max effort | Jul 16, 2026Run date | 56active users | Source for Kimi K3 (opens in a new tab) |
| 20 | Gemini 3.5 FlashListed as gemini-3.5-flash_high | 92.8%Standard error ±1.6 pts | High effort | May 22, 2026Run date | 3active users | Source for Gemini 3.5 Flash (opens in a new tab) |
| 21 | Grok 4.7Listed as grok-4.7_xhigh | 92.7%Standard error ±1.6 pts | XHigh effort | Sep 22, 2026Run date | 131active users | Source for Grok 4.7 (opens in a new tab) |
| 22 | Qwen3.8 MaxListed as qwen3.8-max_xhigh | 92.7%Standard error ±1.7 pts | XHigh effort | Aug 4, 2026Run date | 16active users | Source for Qwen3.8 Max (opens in a new tab) |
| 23 | Gemini 3 Pro PreviewListed as gemini-3-pro-preview | 92.6%Standard error ±1.7 pts | Not reported | Nov 19, 2025Run date | 1active users | Source for Gemini 3 Pro Preview (opens in a new tab) |
| 24 | Qwen3.8 Max 0902Listed as qwen3.8-max-0902_xhigh | 92.3%Standard error ±1.7 pts | XHigh effort | Sep 2, 2026Run date | 1active users | Source for Qwen3.8 Max 0902 (opens in a new tab) |
| 25 | GLM-5.2Listed as glm-5.2_max | 91.9%Standard error ±1.6 pts | Max effort | Jun 24, 2026Run date | 13active users | Source for GLM-5.2 (opens in a new tab) |
| 26 | DeepSeek V4 Pro 0813Listed as deepseek-v4-pro-0813_max | 91.7%Standard error ±1.5 pts | Max effort | Aug 18, 2026Run date | 6active users | Source for DeepSeek V4 Pro 0813 (opens in a new tab) |
| 27 | GPT-5.6 LunaListed as gpt-5.6-luna_max | 91.6%Standard error ±1.7 pts | Max effort | Jul 9, 2026Run date | 158active users | Source for GPT-5.6 Luna (opens in a new tab) |
| 28 | GPT-5.2Listed as gpt-5.2-2025-12-11_xhigh | 91.4%Standard error ±1.8 pts | XHigh effort | Dec 13, 2025Run date | No Tokscale usage in the last 30 days | Source for GPT-5.2 (opens in a new tab) |
| 29 | DeepSeek V4 Flash 0731Listed as deepseek-v4-flash-0731_max | 91.0%Standard error ±1.8 pts | Max effort | Aug 2, 2026Run date | 18active users | Source for DeepSeek V4 Flash 0731 (opens in a new tab) |
| 29 | Claude Opus 4.8Listed as claude-opus-4-8_max | 91.0%Standard error ±1.9 pts | Max effort | Jun 7, 2026Run date | 222active users | Source for Claude Opus 4.8 (opens in a new tab) |
| 31 | GLM-5.3Listed as glm-5.3_max | 90.9%Standard error ±1.6 pts | Max effort | Aug 24, 2026Run date | 139active users | Source for GLM-5.3 (opens in a new tab) |
| 31 | MiniMax M3Listed as MiniMax-M3 | 90.9%Standard error ±2.0 pts | Not reported | Aug 10, 2026Run date | 26active users | Source for MiniMax M3 (opens in a new tab) |
| 31 | Qwen3.7 MaxListed as qwen3.7-max | 90.9%Standard error ±2.0 pts | Not reported | Aug 7, 2026Run date | 3active users | Source for Qwen3.7 Max (opens in a new tab) |
| 31 | DeepSeek V4 ProListed as deepseek-v4-pro_high | 90.9%Standard error ±2.0 pts | High effort | Aug 6, 2026Run date | 51active users | Source for DeepSeek V4 Pro (opens in a new tab) |
| 35 | Kimi K2.6Listed as kimi-k2.6 | 90.8%Standard error ±1.7 pts | Not reported | May 1, 2026Run date | 4active users | Source for Kimi K2.6 (opens in a new tab) |
| 36 | Claude Opus 5.5Listed as claude-opus-5-5_max | 90.6%Standard error ±1.9 pts | Max effort | Sep 22, 2026Run date | 709active users | Source for Claude Opus 5.5 (opens in a new tab) |
| 37 | Claude Sonnet 5Listed as claude-sonnet-5_xhigh | 90.5%Standard error ±1.8 pts | XHigh effort | Jul 1, 2026Run date | 138active users | Source for Claude Sonnet 5 (opens in a new tab) |
| 37 | Claude Opus 4.6Listed as claude-opus-4-6_32K | 90.5%Standard error ±1.7 pts | 32K thinking budget | Feb 6, 2026Run date | 52active users | Source for Claude Opus 4.6 (opens in a new tab) |
| 39 | GPT-6 LunaListed as gpt-6-luna_max | 90.5%Standard error ±1.9 pts | Max effort | Sep 22, 2026Run date | 433active users | Source for GPT-6 Luna (opens in a new tab) |
| 40 | GLM-5.3 FlashListed as glm-5.3-flash_max | 90.2%Standard error ±1.7 pts | Max effort | Aug 26, 2026Run date | 170active users | Source for GLM-5.3 Flash (opens in a new tab) |
| 40 | Claude Opus 4.7Listed as claude-opus-4-7_xhigh | 90.2%Standard error ±1.8 pts | XHigh effort | Apr 17, 2026Run date | 27active users | Source for Claude Opus 4.7 (opens in a new tab) |
| 42 | GLM-5.1Listed as glm-5.1 | 89.9%Standard error ±2.1 pts | Not reported | Aug 10, 2026Run date | 5active users | Source for GLM-5.1 (opens in a new tab) |
| 43 | DeepSeek V4.1 FlashListed as deepseek-v4.1-flash_max | 89.8%Standard error ±1.7 pts | Max effort | Oct 9, 2026Run date | 225active users | Source for DeepSeek V4.1 Flash (opens in a new tab) |
| 44 | muse-sparkNot matched to a Tokscale model | 89.8%Standard error ±2.2 pts | Not reported | Apr 8, 2026Run date | Not matched to a Tokscale model | Source for muse-spark (opens in a new tab) |
| 45 | Claude Haiku 5.5Listed as claude-haiku-5-5_max | 89.6%Standard error ±2.0 pts | Max effort | Oct 9, 2026Run date | 477active users | Source for Claude Haiku 5.5 (opens in a new tab) |
| 46 | Gemini 3 Flash PreviewListed as gemini-3-flash-preview_high | 89.4%Standard error ±2.2 pts | High effort | Aug 6, 2026Run date | 4active users | Source for Gemini 3 Flash Preview (opens in a new tab) |
| 47 | Grok 4.20 0309 ReasoningListed as grok-4.20-0309-reasoning | 89.3%Standard error ±1.9 pts | Not reported | Jul 13, 2026Run date | 1active users | Source for Grok 4.20 0309 Reasoning (opens in a new tab) |
| 48 | Grok 4.3Listed as grok-4.3_high | 88.8%Standard error ±2.0 pts | High effort | Jun 17, 2026Run date | No Tokscale usage in the last 30 days | Source for Grok 4.3 (opens in a new tab) |
| 49 | Muse Spark OpenListed as Muse-Spark-Open-816B-A42B_xhigh | 88.6%Standard error ±1.9 pts | XHigh effort | Oct 15, 2026Run date | 2active users | Source for Muse Spark Open (opens in a new tab) |
| 50 | Inkling SmallListed as inkling-small_xhigh | 88.5%Standard error ±1.9 pts | XHigh effort | Aug 14, 2026Run date | No Tokscale usage in the last 30 days | Source for Inkling Small (opens in a new tab) |
Showing 50 of 223 models
Score by release date
Each dot is one model's best result on GPQA Diamond, placed at the model's release date. The dashed line is the best score among models released up to that date.
- Used on Tokscale in the last 30 days
- No Tokscale usage in the last 30 days
- Best score so far
Gemini 4 Argon: 96.4% · released Sep 30, 2026 · High effort · 82 active users on Tokscale in the last 30 days
151 models, released Feb 26, 2024 to Oct 15, 2026. Highest: Gemini 4 Argon, 96.4%, released Sep 30, 2026. Use the left and right arrow keys to read each model in release order.0%20%40%60%80%100%
Release dates are from Epoch AI; 1 comes from models.dev where Epoch AI lists none.
Show data
| Model | Released | Score | Tokscale users, 30 days |
|---|---|---|---|
| Muse Spark Open | Oct 15, 2026 | 88.6%XHigh effort | 2 |
| Claude Haiku 5.5 | Oct 7, 2026 | 89.6%Max effort | 477 |
| Mistral Large 4 | Oct 6, 2026 | 88.1%High effort | 4 |
| Gemini 4 Argon | Sep 30, 2026 | 96.4%High effort | 82 |
| GPT-6.1 Sol | Sep 29, 2026 | 95.4%Max effort | 660 |
| Claude Sonnet 5.5 | Sep 28, 2026 | 95.6%Max effort | 555 |
| GPT-6 Luna | Sep 22, 2026 | 90.5%Max effort | 433 |
| Claude Opus 5.5 | Sep 22, 2026 | 90.6%Max effort | 709 |
| GPT-6 Sol | Sep 22, 2026 | 94.3%Max effort | 246 |
| Grok 4.7 | Sep 21, 2026 | 92.7%XHigh effort | 131 |
| DeepSeek V4.1 Flash | Sep 9, 2026 | 89.8%Max effort | 225 |
| GPT-6 Astra | Sep 3, 2026 | 95.8%Max effort | 563 |
| Gemini 3.8 Flash | Sep 2, 2026 | 95.4%High effort | 66 |
| Qwen3.8 Max 0902 | Sep 1, 2026 | 92.3%XHigh effort | 1 |
| GLM-5.3 Flash | Aug 20, 2026 | 90.2%Max effort | 170 |
| GLM-5.3 | Aug 14, 2026 | 90.9%Max effort | 139 |
| DeepSeek V4 Pro 0813 | Aug 13, 2026 | 91.7%Max effort | 6 |
| Gemini 3.7 Flash | Aug 13, 2026 | 94.8%High effort | 4 |
| Grok 4.6 | Aug 12, 2026 | 94.0%High effort | 41 |
| Qwen3.8 Max | Aug 2, 2026 | 92.7%XHigh effort | 16 |
| DeepSeek V4 Flash 0731 | Jul 31, 2026 | 91.0%Max effort | 18 |
| Qwen3.7 Flash | Jul 27, 2026 | 82.3% | No Tokscale usage in the last 30 days |
| Claude Opus 5 | Jul 24, 2026 | 93.9%Max effort | 163 |
| Gemini 3.5 Flash Lite | Jul 21, 2026 | 83.3%High effort | 13 |
| Gemini 3.6 Flash | Jul 21, 2026 | 94.1%High effort | 3 |
| Kimi K3 | Jul 16, 2026 | 93.1%Max effort | 56 |
| Inkling | Jul 15, 2026 | 88.3%XHigh effort | No Tokscale usage in the last 30 days |
| Inkling Small | Jul 15, 2026 | 88.5%XHigh effort | No Tokscale usage in the last 30 days |
| GPT-5.6 Luna | Jul 9, 2026 | 91.6%Max effort | 158 |
| GPT-5.6 Terra | Jul 9, 2026 | 93.3%Max effort | 120 |
| GPT-5.6 Sol | Jul 9, 2026 | 93.5%Max effort | 217 |
| Grok 4.5 | Jul 8, 2026 | 93.4%High effort | 9 |
| Claude Sonnet 5 | Jun 30, 2026 | 90.5%XHigh effort | 138 |
| GLM-5.2 | Jun 16, 2026 | 91.9%Max effort | 13 |
| Kimi K2.7 Code | Jun 12, 2026 | 87.9% | 17 |
| Claude Fable 5 | Jun 9, 2026 | 85.9%Max effort | 41 |
| Qwen3.7 Plus | Jun 2, 2026 | 87.9% | 3 |
| MiniMax M3 | Jun 1, 2026 | 90.9% | 26 |
| Claude Opus 4.8 | May 28, 2026 | 91.0%Max effort | 222 |
| Qwen3.7 Max | May 19, 2026 | 90.9% | 3 |
| Gemini 3.5 Flash | May 19, 2026 | 92.8%High effort | 3 |
| GPT-5.5 Instant | May 5, 2026 | 82.5% | No Tokscale usage in the last 30 days |
| Qwen3.6 Flash | Apr 27, 2026 | 83.3% | No Tokscale usage in the last 30 days |
| DeepSeek V4 Pro | Apr 24, 2026 | 90.9%High effort | 51 |
| GPT-5.5 Pro | Apr 23, 2026 | 93.9%XHigh effort | 1 |
| GPT-5.5 | Apr 23, 2026 | 94.0%XHigh effort | 30 |
| Qwen3.6 27B | Apr 22, 2026 | 85.9% | No Tokscale usage in the last 30 days |
| Qwen3.6 Max Preview | Apr 20, 2026 | 87.4% | No Tokscale usage in the last 30 days |
| Kimi K2.6 | Apr 20, 2026 | 90.8% | 4 |
| Grok 4.3 | Apr 17, 2026 | 88.8%High effort | No Tokscale usage in the last 30 days |
| Claude Opus 4.7 | Apr 16, 2026 | 90.2%XHigh effort | 27 |
| Qwen3.6 35B A3B | Apr 14, 2026 | 84.8%No reasoning | 1 |
| GLM-5.1 | Apr 7, 2026 | 89.9% | 5 |
| Gemma 4 26B A4B IT | Apr 2, 2026 | 73.2%Minimal effort | 1 |
| Gemma 4 31B IT | Apr 2, 2026 | 75.8%Minimal effort | 3 |
| Qwen3.6 Plus | Mar 31, 2026 | 88.4% | 1 |
| GPT-5.4 Nano | Mar 17, 2026 | 78.5%High effort | 4 |
| GPT-5.4 Mini | Mar 17, 2026 | 86.9%XHigh effort | 1 |
| GPT-5.4 | Mar 5, 2026 | 93.3%XHigh effort | 1 |
| GPT-5.4 Pro | Mar 5, 2026 | 94.6%XHigh effort | No Tokscale usage in the last 30 days |
| Gemini 3.1 Flash Lite | Mar 3, 2026 | 81.8%High effort | 5 |
| Qwen3.5 Flash | Feb 25, 2026 | 82.3% | No Tokscale usage in the last 30 days |
| Qwen3.5 9B | Feb 24, 2026 | 79.0% | 2 |
| Qwen3.5 35B A3B | Feb 24, 2026 | 83.5% | No Tokscale usage in the last 30 days |
| Gemini 3.1 Pro Preview | Feb 19, 2026 | 94.4%High effort | 5 |
| Claude Sonnet 4.6 | Feb 17, 2026 | 87.4%32K thinking budget | 55 |
| Grok 4.20 0309 Reasoning | Feb 17, 2026 | 89.3% | 1 |
| Qwen3.5 Plus | Feb 16, 2026 | 84.8% | No Tokscale usage in the last 30 days |
| Qwen3.5 397B A17B | Feb 13, 2026 | 86.4%No reasoning | No Tokscale usage in the last 30 days |
| GLM-5 | Feb 11, 2026 | 87.8% | 2 |
| Claude Opus 4.6 | Feb 5, 2026 | 90.5%32K thinking budget | 52 |
| Kimi K2p5 | Jan 27, 2026 | 87.6% | No Tokscale usage in the last 30 days |
| GLM-4.7 Flash | Jan 19, 2026models.dev | 60.5% | 4 |
| GLM-4.7 | Dec 22, 2025 | 83.3% | 4 |
| Gemini 3 Flash Preview | Dec 17, 2025 | 89.4%High effort | 4 |
| GPT-5.2 | Dec 11, 2025 | 91.4%XHigh effort | No Tokscale usage in the last 30 days |
| DeepSeek Chat | Dec 1, 2025 | 71.2% | 2 |
| DeepSeek Reasoner | Dec 1, 2025 | 83.4% | No Tokscale usage in the last 30 days |
| Claude Opus 4.5 | Nov 24, 2025 | 86.0%32K thinking budget | 4 |
| Gemini 3 Pro Preview | Nov 18, 2025 | 92.6% | 1 |
| GPT-5.1 | Nov 13, 2025 | 87.6%High effort | No Tokscale usage in the last 30 days |
| Kimi K2 Thinking | Nov 6, 2025 | 84.2% | No Tokscale usage in the last 30 days |
| Claude Haiku 4.5 | Oct 15, 2025 | 71.2%32K thinking budget | 102 |
| Claude Sonnet 4.5 | Sep 29, 2025 | 82.3%59K thinking budget | 23 |
| Qwen3 Max | Sep 24, 2025 | 72.6% | No Tokscale usage in the last 30 days |
| Magistral Small 2509 | Sep 18, 2025 | 47.6% | No Tokscale usage in the last 30 days |
| Seed OSS 36B Instruct | Aug 20, 2025 | 71.5% | No Tokscale usage in the last 30 days |
| GPT-5 Nano | Aug 7, 2025 | 69.4%High effort | No Tokscale usage in the last 30 days |
| GPT-5 Mini | Aug 7, 2025 | 75.0%High effort | 1 |
| GPT-5 2025 08 07 | Aug 7, 2025 | 86.2%High effort | No Tokscale usage in the last 30 days |
| GPT OSS 20B | Aug 5, 2025 | 60.8%Medium effort | 2 |
| GPT OSS 120B | Aug 5, 2025 | 75.8%High effort | 6 |
| Claude Opus 4.1 | Aug 5, 2025 | 77.3%16K thinking budget | No Tokscale usage in the last 30 days |
| Grok 4 0709 | Jul 9, 2025 | 87.0% | No Tokscale usage in the last 30 days |
| Gemini 2.5 Pro | Jun 17, 2025 | 85.3% | 1 |
| Magistral Small 2506 | Jun 10, 2025 | 56.1% | No Tokscale usage in the last 30 days |
| Gemini 2.5 Pro Preview 06 05 | Jun 5, 2025 | 84.8% | No Tokscale usage in the last 30 days |
| DeepSeek R1 0528 Qwen3 8B | May 28, 2025 | 9.3% | No Tokscale usage in the last 30 days |
| DeepSeek R1 0528 | May 28, 2025 | 76.3% | No Tokscale usage in the last 30 days |
| Claude Opus 4 | May 22, 2025 | 76.3%16K thinking budget | No Tokscale usage in the last 30 days |
| Claude Sonnet 4 | May 22, 2025 | 79.2%59K thinking budget | No Tokscale usage in the last 30 days |
| Mistral Medium 2505 | May 7, 2025 | 59.5% | No Tokscale usage in the last 30 days |
| Gemini 2.5 Pro Preview 05 06 | May 6, 2025 | 66.7% | No Tokscale usage in the last 30 days |
| Qwen3 4B | Apr 29, 2025 | 52.3% | 2 |
| Qwen3 30B A3B | Apr 29, 2025 | 61.7% | No Tokscale usage in the last 30 days |
| Qwen3 14B | Apr 29, 2025 | 63.8% | No Tokscale usage in the last 30 days |
| Qwen3 32B | Apr 29, 2025 | 65.7% | No Tokscale usage in the last 30 days |
| Qwen3 235B A22B | Apr 29, 2025 | 70.7% | No Tokscale usage in the last 30 days |
| Qwen3 8B | Apr 28, 2025 | 56.8% | 1 |
| o4-mini | Apr 16, 2025 | 79.6%High effort | No Tokscale usage in the last 30 days |
| o3 | Apr 16, 2025 | 81.8%High effort | No Tokscale usage in the last 30 days |
| GPT-4.1 Nano | Apr 14, 2025 | 48.9% | No Tokscale usage in the last 30 days |
| GPT-4.1 Mini 2025 04 14 | Apr 14, 2025 | 65.8% | 5 |
| GPT-4.1 2025 04 14 | Apr 14, 2025 | 66.9% | 5 |
| Grok 3 Beta | Apr 9, 2025 | 75.8% | No Tokscale usage in the last 30 days |
| Grok 3 Mini | Apr 9, 2025 | 76.3%Low effort | No Tokscale usage in the last 30 days |
| QwQ Plus | Apr 8, 2025 | 65.4% | No Tokscale usage in the last 30 days |
| Gemini 2.5 Pro Exp 03 25 | Mar 25, 2025 | 83.8% | No Tokscale usage in the last 30 days |
| DeepSeek V3 0324 | Mar 24, 2025 | 67.6% | No Tokscale usage in the last 30 days |
| Gemma 3 4B IT | Mar 12, 2025 | 23.2% | No Tokscale usage in the last 30 days |
| Gemma 3 12B IT | Mar 12, 2025 | 39.5% | No Tokscale usage in the last 30 days |
| Gemma 3 27B IT | Mar 12, 2025 | 47.7% | No Tokscale usage in the last 30 days |
| QwQ 32B | Mar 5, 2025 | 65.3% | No Tokscale usage in the last 30 days |
| Claude 3.7 Sonnet | Feb 24, 2025 | 79.7%64K thinking budget | No Tokscale usage in the last 30 days |
| o3-mini | Jan 31, 2025 | 77.0%High effort | No Tokscale usage in the last 30 days |
| Qwen Plus | Jan 25, 2025 | 48.1% | No Tokscale usage in the last 30 days |
| Qwen Max | Jan 25, 2025 | 56.1% | No Tokscale usage in the last 30 days |
| DeepSeek R1 | Jan 20, 2025 | 71.7% | No Tokscale usage in the last 30 days |
| DeepSeek V3 | Dec 26, 2024 | 56.5% | No Tokscale usage in the last 30 days |
| o1 | Dec 17, 2024 | 76.8%High effort | No Tokscale usage in the last 30 days |
| Llama 3.3 70B Instruct | Dec 6, 2024 | 47.4% | No Tokscale usage in the last 30 days |
| GPT-4o 2024 11 20 | Nov 20, 2024 | 47.9% | 1 |
| Mistral Large 2411 | Nov 18, 2024 | 51.3% | No Tokscale usage in the last 30 days |
| Qwen Turbo | Nov 1, 2024 | 41.8% | No Tokscale usage in the last 30 days |
| Claude 3.5 Haiku | Oct 22, 2024 | 38.1% | No Tokscale usage in the last 30 days |
| Ministral 3B | Oct 16, 2024 | 25.3% | No Tokscale usage in the last 30 days |
| Llama 3.2 1B Instruct | Sep 24, 2024 | 23.9% | No Tokscale usage in the last 30 days |
| Llama 3.2 90B Vision Instruct | Sep 24, 2024 | 41.0% | No Tokscale usage in the last 30 days |
| GPT-4o 2024 08 06 | Aug 6, 2024 | 49.2% | No Tokscale usage in the last 30 days |
| Mistral Large 2407 | Jul 24, 2024 | 49.0% | No Tokscale usage in the last 30 days |
| Llama 3.1 8B Instruct | Jul 23, 2024 | 27.0% | No Tokscale usage in the last 30 days |
| Llama 3.1 70B Instruct | Jul 23, 2024 | 44.2% | No Tokscale usage in the last 30 days |
| Mistral Nemo | Jul 18, 2024 | 29.9% | No Tokscale usage in the last 30 days |
| GPT-4o Mini | Jul 18, 2024 | 37.7% | 1 |
| Gemma 2 27B IT | Jun 24, 2024 | 36.5% | No Tokscale usage in the last 30 days |
| Mistral 7B Instruct V0.3 | May 27, 2024 | 15.2% | No Tokscale usage in the last 30 days |
| GPT-4o 2024 05 13 | May 13, 2024 | 48.9% | No Tokscale usage in the last 30 days |
| Wizardlm 2 8x22b | Apr 15, 2024 | 43.4% | No Tokscale usage in the last 30 days |
| GPT-4 Turbo | Apr 9, 2024 | 46.6% | No Tokscale usage in the last 30 days |
| Claude 3 Haiku | Mar 7, 2024 | 36.3% | No Tokscale usage in the last 30 days |
| Mistral Large | Feb 26, 2024 | 38.8% | No Tokscale usage in the last 30 days |
Method and matching
How results are run
- Run by Epoch AI at each reasoning effort.
Matching to Tokscale models
- 250 of 324 results are matched to 151 Tokscale models: 168 by exact id, 82 by normalized name.
- No match is marked as reviewed by hand yet.
- 74 results from 72 models are not matched. They stay in the table without a Tokscale link.
- Reasoning effort, thinking budget and agent are conditions of a result. They never change which model a result belongs to.
License and attribution
- CC BY 4.0 (opens in a new tab), retrieved Oct 21, 2026.
- Epoch AI, ‘Capabilities & benchmarking’, epoch.ai/benchmarks (CC BY 4.0)
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.