MirrorCode
Re-run by Epoch AINo description yet: we only describe an eval set when we can do it correctly.
- Kind
- Run independently by Epoch AI, with public logs. Our own re-runs are not built yet.
- Score shown
- Epoch's column "Best score (across scorers)", as a percentage, higher is better; random guessing scores about 0.0%. Error bars as the source gives them.
- Results
- 47 in all: 5 for open-weight models, 42 for closed ones. 4 linked to a model page here.
Best open-weight result: deepseek-ai/DeepSeek-V4.1-Flash, 64.4%, number 12 of 47 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | claude-sonnet-5-5_maxAnthropic | closed | effort max | 97.8%±1.5 | 5 Oct 2026 | Epoch AI log |
| 2 | claude-opus-5-5_maxAnthropic | closed | effort max | 95.6%±2.6 | 30 Oct 2026 | Epoch AI |
| 3 | claude-fable-5-1_maxAnthropic | closed | effort max | 93.3%±3.9 | 9 Sep 2026 | Epoch AI log |
| 4 | claude-opus-5_maxAnthropic | closed | effort max | 87.8%±7.2 | 20 Oct 2026 | Epoch AI log |
| 5 | claude-opus-5-5_highAnthropic | closed | effort high | 78.9%±7.9 | 22 Sep 2026 | Epoch AI |
| 6 | claude-haiku-5-5_maxAnthropic | closed | effort max | 78.9%±8.0 | 21 Oct 2026 | Epoch AI log |
| 7 | gpt-6.1-terra_maxOpenAI | closed | effort max | 76.7%±9.3 | 30 Oct 2026 | Epoch AI |
| 8 | claude-fable-5-1_highAnthropic | closed | effort high | 73.3%±9.2 | 10 Sep 2026 | Epoch AI |
| 9 | claude-fable-5_maxAnthropic | closed | effort max | 70.0%±10.4 | 30 Oct 2026 | Epoch AI |
| 10 | gpt-6.2-astra_maxOpenAI | closed | effort max | 67.8%±11.7 | 27 Oct 2026 | Epoch AI |
| 11 | claude-opus-4-8_maxAnthropic | closed | effort max | 65.6%±10.8 | 21 Oct 2026 | Epoch AI log |
| 12 | deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_max, DeepSeek | open | effort max | 64.4%±11.4 | 21 Oct 2026 | Epoch AI |
| 13 | claude-fable-5_highAnthropic | closed | effort high | 63.9%±10.4 | 10 Aug 2026 | Epoch AI |
| 14 | claude-sonnet-5_maxAnthropic | closed | effort max | 63.3%±10.7 | 21 Oct 2026 | Epoch AI |
| 15 | gpt-6.1-sol_maxOpenAI | closed | effort max | 61.1%±11.3 | 5 Oct 2026 | Epoch AI log |
| 16 | gpt-6-astra_maxOpenAI | closed | effort max | 60.0%±11.8 | 29 Oct 2026 | Epoch AI log |
| 17 | gemini-4-argon_highGoogle DeepMind | closed | effort high | 58.9%±11.7 | 14 Oct 2026 | Epoch AI log |
| 18 | grok-4.7_xhighxAI | closed | effort xhigh | 54.4%±11.4 | 29 Oct 2026 | Epoch AI |
| 19 | moonshotai/Kimi-K3Epoch: kimi-k3_max, Moonshot | open | effort max | 54.4%±11.4 | 3 Sep 2026 | Epoch AI log |
| 20 | muse-spark-1.3_maxMeta AI | closed | effort max | 54.4%±10.8 | 20 Oct 2026 | Epoch AI |
| 21 | qwen3.8-max_xhighAlibaba | closed | effort xhigh | 52.2%±10.9 | 21 Oct 2026 | Epoch AI log |
| 22 | gpt-6-sol_maxOpenAI | closed | effort max | 48.9%±11.2 | 29 Oct 2026 | Epoch AI |
| 23 | qwen3.8-max-0902_xhighAlibaba | closed | effort xhigh | 47.8%±10.3 | 21 Oct 2026 | Epoch AI log |
| 24 | gpt-6-astra_highOpenAI | closed | effort high | 46.7%±11.9 | 30 Aug 2026 | Epoch AI |
| 25 | muse-wonder-1.0_maxMeta AI | closed | effort max | 45.6%±10.6 | 14 Oct 2026 | Epoch AI log |
| 26 | claude-opus-4-7_maxAnthropic | closed | effort max | 43.3%±9.6 | 30 Oct 2026 | Epoch AI |
| 27 | gemini-3.7-flash_highGoogle DeepMind | closed | effort high | 41.1%±12.2 | 21 Oct 2026 | Epoch AI log |
| 28 | gpt-5.6-terra_maxOpenAI | closed | effort max | 41.1%±12.4 | 20 Oct 2026 | Epoch AI log |
| 29 | grok-4.6_xhighxAI | closed | effort xhigh | 41.1%±11.0 | 21 Oct 2026 | Epoch AI log |
| 30 | mistral-large-4-0_highMistral AIno page match | open | effort high | 40.0%±10.6 | 2 Nov 2026 | Epoch AI |
| 31 | gemini-3.8-flash_highGoogle DeepMind | closed | effort high | 38.9%±11.9 | 2 Sep 2026 | Epoch AI log |
| 32 | zai-org/GLM-5.2Epoch: glm-5.2_max, Z.ai (Zhipu AI) | open | effort max | 36.7%±11.5 | 21 Oct 2026 | Epoch AI |
| 33 | claude-opus-4-6_maxAnthropic | closed | effort max | 33.3%±10.0 | 21 Oct 2026 | Epoch AI log |
| 34 | deepseek-ai/DeepSeek-V4-Pro-0813Epoch: deepseek-v4-pro-0813_max, DeepSeek | open | effort max | 33.3%±10.0 | 20 Oct 2026 | Epoch AI log |
| 35 | gpt-5.2-2025-12-11_xhighOpenAI | closed | effort xhigh | 32.2%±10.0 | 21 Oct 2026 | Epoch AI |
| 36 | grok-4.5_highxAI | closed | effort high | 32.2%±10.7 | 21 Oct 2026 | Epoch AI log |
| 37 | claude-opus-4-7_highAnthropic | closed | effort high | 31.1%±9.0 | 12 Aug 2026 | Epoch AI |
| 38 | gpt-6-luna_maxOpenAI | closed | effort max | 26.7%±9.0 | 21 Oct 2026 | Epoch AI log |
| 39 | gemini-3.6-flash_highGoogle DeepMind | closed | effort high | 23.3%±9.3 | 21 Oct 2026 | Epoch AI |
| 40 | gpt-5.4-2026-03-05_xhighOpenAI | closed | effort xhigh | 23.3%±8.1 | 29 Oct 2026 | Epoch AI |
| 41 | gpt-5.5_xhighOpenAI | closed | effort xhigh | 23.3%±10.1 | 29 Oct 2026 | Epoch AI |
| 42 | gpt-5.6-luna_maxOpenAI | closed | effort max | 21.1%±9.8 | 21 Oct 2026 | Epoch AI |
| 43 | gpt-5.6-sol_highOpenAI | closed | effort high | 20.0%±9.2 | 10 Aug 2026 | Epoch AI |
| 44 | gpt-5.4-2026-03-05_highOpenAI | closed | effort high | 15.6%±7.7 | 10 Aug 2026 | Epoch AI |
| 45 | gpt-5.5_highOpenAI | closed | effort high | 10.0%±6.0 | 10 Aug 2026 | Epoch AI |
| 46 | gemini-3.1-pro-preview_highGoogle DeepMind | closed | effort high | 8.9%±4.6 | 10 Aug 2026 | Epoch AI |
| 47 | gpt-5-2025-08-07_highOpenAI | closed | effort high | 4.4%±3.4 | 21 Oct 2026 | Epoch AI log |
Open-weight rows are shaded. Dates are when the result was run or evaluated, where the data says. Data from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Epoch AI. Method