Mystery Game Puzzles
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 9.2%. Error bars as the source gives them.
- Results
- 148 in all: 29 for open-weight models, 112 for closed ones, 7 where Epoch does not say. 26 linked to a model page here.
Best open-weight result: deepseek-ai/DeepSeek-V4.1-Flash-1017, 49.0%, number 19 of 148 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | gpt-6.2-astra_maxOpenAI | closed | effort max | 95.0%±2.2 | 27 Oct 2026 | Epoch AI |
| 2 | claude-fable-5-2_maxAnthropic | closed | effort max | 90.0%±3.0 | 6 Nov 2026 | Epoch AI |
| 3 | gpt-6-astra_maxOpenAI | closed | effort max | 84.0%±3.7 | 30 Aug 2026 | Epoch AI |
| 4 | gpt-6.1-sol_maxOpenAI | closed | effort max | 80.0%±4.0 | 29 Sep 2026 | Epoch AI |
| 5 | gpt-6.1-terra_maxOpenAI | closed | effort max | 74.0%±4.4 | 26 Oct 2026 | Epoch AI |
| 6 | gemini-4-carbon_highGoogle DeepMind | closed | effort high | 73.0%±4.5 | 2 Nov 2026 | Epoch AI |
| 7 | claude-opus-5-5_maxAnthropic | closed | effort max | 71.0%±4.6 | 29 Sep 2026 | Epoch AI |
| 8 | claude-sonnet-5-5_maxAnthropic | closed | effort max | 65.0%±4.8 | 29 Sep 2026 | Epoch AI |
| 9 | gemini-4-hydrogen_xhighGoogle DeepMind | closed | effort xhigh | 65.0%±4.8 | 13 Nov 2026 | Epoch AI |
| 10 | gemini-4-argon_highGoogle DeepMind | closed | effort high | 64.0%±4.8 | 13 Oct 2026 | Epoch AI |
| 11 | claude-opus-5_maxAnthropic | closed | effort max | 59.0%±4.9 | 25 Jul 2026 | Epoch AI |
| 12 | claude-fable-5-1_maxAnthropic | closed | effort max | 58.0%±5.0 | 1 Sep 2026 | Epoch AI |
| 13 | gpt-5.6-sol_maxOpenAI | closed | effort max | 58.0%±5.0 | 28 Jul 2026 | Epoch AI |
| 14 | gpt-5.5_xhighOpenAI | closed | effort xhigh | 56.0%±5.0 | 24 Jul 2026 | Epoch AI |
| 15 | gpt-6-sol_maxOpenAI | closed | effort max | 56.0%±5.0 | 22 Sep 2026 | Epoch AI |
| 16 | claude-fable-5_maxAnthropic | closed | effort max | 52.0%±5.0 | 31 Jul 2026 | Epoch AI |
| 17 | gpt-5.5_highOpenAI | closed | effort high | 52.0%±5.0 | 27 Jul 2026 | Epoch AI |
| 18 | muse-wonder-1.0_maxMeta AI | closed | effort max | 51.0%±5.0 | 14 Oct 2026 | Epoch AI |
| 19 | deepseek-ai/DeepSeek-V4.1-Flash-1017Epoch: deepseek-v4.1-flash-1017_max, DeepSeek | open | effort max | 49.0%±5.0 | 19 Oct 2026 | Epoch AI |
| 20 | gemini-3.8-flash_highGoogle DeepMind | closed | effort high | 47.0%±5.0 | 2 Sep 2026 | Epoch AI |
| 21 | deepseek-ai/DeepSeek-V4-Pro-0813Epoch: deepseek-v4-pro-0813_max, DeepSeek | open | effort max | 43.0%±5.0 | 19 Aug 2026 | Epoch AI |
| 22 | deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_max, DeepSeek | open | effort max | 43.0%±5.0 | 8 Oct 2026 | Epoch AI |
| 23 | moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro-together_max, Moonshot | open | effort max | 42.0%±5.0 | 6 Nov 2026 | Epoch AI |
| 24 | qwen3.8-max_xhighAlibaba | closed | effort xhigh | 38.0%±4.9 | 5 Aug 2026 | Epoch AI |
| 25 | claude-opus-5Anthropic | closed | default | 37.0%±4.9 | 6 Aug 2026 | Epoch AI log |
| 26 | gemini-3.7-flash_highGoogle DeepMind | closed | effort high | 37.0%±4.9 | 14 Aug 2026 | Epoch AI |
| 27 | gpt-5.4-2026-03-05_xhighOpenAI | closed | effort xhigh | 37.0%±4.9 | 24 Jul 2026 | Epoch AI |
| 28 | claude-opus-4-8_maxAnthropic | closed | effort max | 36.0%±4.8 | 25 Jul 2026 | Epoch AI |
| 29 | claude-sonnet-5_maxAnthropic | closed | effort max | 35.0%±4.8 | 28 Jul 2026 | Epoch AI |
| 30 | gpt-5.6-terra_maxOpenAI | closed | effort max | 35.0%±4.8 | 28 Jul 2026 | Epoch AI |
| 31 | deepseek-ai/DeepSeek-V4-Flash-0731Epoch: deepseek-v4-flash-0731_max, DeepSeek | open | effort max | 34.0%±4.8 | 5 Aug 2026 | Epoch AI |
| 32 | gemini-3.1-pro-preview_highGoogle DeepMind | closed | effort high | 34.0%±4.8 | 27 Jul 2026 | Epoch AI |
| 33 | grok-4.6_xhighxAI | closed | effort xhigh | 34.0%±4.8 | 14 Aug 2026 | Epoch AI |
| 34 | zai-org/GLM-5.3Epoch: glm-5.3_max, Z.ai (Zhipu AI) | open | effort max | 33.0%±4.7 | 30 Aug 2026 | Epoch AI |
| 35 | gpt-5.6-sol_noneOpenAI | closed | effort none | 33.0%±4.7 | 27 Aug 2026 | Epoch AI log |
| 36 | gemini-3.1-pro-preview_mediumGoogle DeepMind | closed | effort medium | 32.0%±4.7 | 6 Aug 2026 | Epoch AI log |
| 37 | gemini-3.5-flash_highGoogle DeepMind | closed | effort high | 32.0%±4.7 | 27 Jul 2026 | Epoch AI |
| 38 | qwen3.7-maxAlibaba | closed | default | 32.0%±4.7 | 28 Jul 2026 | Epoch AI |
| 39 | claude-opus-4-8_xhighAnthropic | closed | effort xhigh | 31.0%±4.6 | 26 Jul 2026 | Epoch AI |
| 40 | claude-haiku-5-5_maxAnthropic | closed | effort max | 30.0%±4.6 | 7 Oct 2026 | Epoch AI |
| 41 | gemini-3.6-flash_highGoogle DeepMind | closed | effort high | 30.0%±4.6 | 5 Aug 2026 | Epoch AI |
| 42 | gemini-3.1-pro-preview_lowGoogle DeepMind | closed | effort low | 29.0%±4.6 | 6 Aug 2026 | Epoch AI log |
| 43 | grok-4.7_xhighxAI | closed | effort xhigh | 29.0%±4.6 | 22 Sep 2026 | Epoch AI |
| 44 | o3-2025-04-16_highOpenAI | closed | effort high | 29.0%±4.6 | 27 Aug 2026 | Epoch AI log |
| 45 | claude-opus-4-7_maxAnthropic | closed | effort max | 28.0%±4.5 | 25 Jul 2026 | Epoch AI |
| 46 | gemini-3.5-flash_lowGoogle DeepMind | closed | effort low | 28.0%±4.5 | 5 Aug 2026 | Epoch AI log |
| 47 | gpt-5.4-2026-03-05_mediumOpenAI | closed | effort medium | 28.0%±4.5 | 28 Aug 2026 | Epoch AI log |
| 48 | gpt-5.5_lowOpenAI | closed | effort low | 28.0%±4.5 | 28 Aug 2026 | Epoch AI log |
| 49 | gemini-3-flash-preview_lowGoogle DeepMind | closed | effort low | 26.0%±4.4 | 5 Aug 2026 | Epoch AI log |
| 50 | gpt-5.6-sol_lowOpenAI | closed | effort low | 26.0%±4.4 | 27 Aug 2026 | Epoch AI log |
| 51 | moonshotai/Kimi-K3Epoch: kimi-k3_max, Moonshot | open | effort max | 26.0%±4.4 | 29 Jul 2026 | Epoch AI |
| 52 | qwen3.7-max_noneAlibaba | closed | effort none | 26.0%±4.4 | 27 Aug 2026 | Epoch AI log |
| 53 | claude-opus-4-6_maxAnthropic | closed | effort max | 25.0%±4.4 | 25 Jul 2026 | Epoch AI |
| 54 | gemini-3-flash-preview_minimalGoogle DeepMind | closed | effort minimal | 25.0%±4.4 | 5 Aug 2026 | Epoch AI log |
| 55 | gemini-3.6-flash_minimalGoogle DeepMind | closed | effort minimal | 25.0%±4.4 | 5 Aug 2026 | Epoch AI log |
| 56 | muse-spark-1.3_maxMeta AI | closed | effort max | 25.0%±4.5 | 18 Sep 2026 | Epoch AI |
| 57 | gemini-3.6-flash_lowGoogle DeepMind | closed | effort low | 23.0%±4.2 | 5 Aug 2026 | Epoch AI log |
| 58 | gpt-5-2025-08-07_highOpenAI | closed | effort high | 23.0%±4.2 | 5 Aug 2026 | Epoch AI |
| 59 | gpt-5.2-2025-12-11_highOpenAI | closed | effort high | 23.0%±4.2 | 6 Aug 2026 | Epoch AI |
| 60 | o3-2025-04-16_mediumOpenAI | closed | effort medium | 23.0%±4.2 | 27 Aug 2026 | Epoch AI log |
| 61 | claude-opus-4-5-20251101_48KAnthropic | closed | default | 22.0%±4.2 | 25 Jul 2026 | Epoch AI |
| 62 | gpt-5.2-2025-12-11_mediumOpenAI | closed | effort medium | 22.0%±4.2 | 28 Aug 2026 | Epoch AI log |
| 63 | Qwen/Qwen3.6-35B-A3BEpoch: qwen3.6-35b-a3b_none, Alibaba | open | effort none | 22.0%±4.2 | 27 Aug 2026 | Epoch AI log |
| 64 | claude-opus-4-1-20250805_24KAnthropic | closed | default | 21.0%±4.1 | 25 Jul 2026 | Epoch AI |
| 65 | gpt-5.6-luna_maxOpenAI | closed | effort max | 21.0%±4.1 | 28 Jul 2026 | Epoch AI |
| 66 | gemini-3-flash-preview_highGoogle DeepMind | closed | effort high | 20.0%±4.0 | 5 Aug 2026 | Epoch AI log |
| 67 | gpt-5.6-luna_noneOpenAI | closed | effort none | 20.0%±4.0 | 30 Aug 2026 | Epoch AI log |
| 68 | mistral-large-4-0-oss_highMistral AIno page match | open | effort high | 20.0%±4.0 | 2 Nov 2026 | Epoch AI |
| 69 | nemotron-3-ultraNvidiano page match | open | default | 20.0%±4.0 | 27 Aug 2026 | Epoch AI log |
| 70 | qwen3.5-flashAlibaba | not stated | default | 20.0%±4.0 | 27 Aug 2026 | Epoch AI log |
| 71 | gemini-3.5-flash-lite_lowGoogle DeepMind | closed | effort low | 19.0%±3.9 | 5 Aug 2026 | Epoch AI log |
| 72 | zai-org/GLM-5.2Epoch: glm-5.2_low, Z.ai (Zhipu AI) | open | effort low | 19.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 73 | gpt-5.1-2025-11-13_lowOpenAI | closed | effort low | 19.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 74 | gpt-5.6-terra_lowOpenAI | closed | effort low | 19.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 75 | qwen3.6-max-preview_noneAlibaba | not stated | effort none | 19.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 76 | zai-org/GLM-5.2Epoch: glm-5.2_minimal, Z.ai (Zhipu AI) | open | effort minimal | 18.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 77 | zai-org/GLM-5.2Epoch: glm-5.2_none, Z.ai (Zhipu AI) | open | effort none | 18.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 78 | gpt-5.5_noneOpenAI | closed | effort none | 18.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 79 | moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshot | open | default | 18.0%±3.9 | 17 Jul 2026 | Epoch AI |
| 80 | Qwen/Qwen3.5-397B-A17BEpoch: qwen3.5-397b-a17b_none, Alibaba | open | effort none | 18.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 81 | qwen3.6-flash_noneAlibaba | not stated | effort none | 18.0%±3.9 | 27 Aug 2026 | Epoch AI log |
| 82 | claude-sonnet-4-5-20250929_48KAnthropic | closed | default | 17.0%±3.8 | 25 Jul 2026 | Epoch AI |
| 83 | deepseek-ai/DeepSeek-V4-ProEpoch: deepseek-v4-pro_none, DeepSeek | open | effort none | 17.0%±3.8 | 27 Aug 2026 | Epoch AI log |
| 84 | gpt-5.4-2026-03-05_lowOpenAI | closed | effort low | 17.0%±3.8 | 27 Aug 2026 | Epoch AI log |
| 85 | gpt-5.6-luna_lowOpenAI | closed | effort low | 17.0%±3.8 | 27 Aug 2026 | Epoch AI log |
| 86 | mistral-large-4-0_highMistral AIno page match | open | effort high | 17.0%±3.8 | 13 Oct 2026 | Epoch AI |
| 87 | Qwen/Qwen3.5-122B-A10BEpoch: qwen3.5-122b-a10b_none, Alibaba | open | effort none | 17.0%±3.8 | 27 Aug 2026 | Epoch AI log |
| 88 | qwen3.5-plus_noneAlibaba | not stated | effort none | 17.0%±3.8 | 27 Aug 2026 | Epoch AI log |
| 89 | qwen3.7-plusAlibaba | closed | default | 17.0%±3.8 | 28 Aug 2026 | Epoch AI log |
| 90 | qwen3.7-plus_noneAlibaba | closed | effort none | 17.0%±3.8 | 27 Aug 2026 | Epoch AI log |
| 91 | claude-sonnet-4-6_lowAnthropic | closed | effort low | 16.0%±3.7 | 6 Aug 2026 | Epoch AI log |
| 92 | claude-sonnet-5Anthropic | closed | default | 16.0%±3.7 | 6 Aug 2026 | Epoch AI log |
| 93 | gpt-5-2025-08-07_lowOpenAI | closed | effort low | 16.0%±3.7 | 27 Aug 2026 | Epoch AI log |
| 94 | gpt-5.1-2025-11-13_mediumOpenAI | closed | effort medium | 16.0%±3.7 | 27 Aug 2026 | Epoch AI log |
| 95 | gpt-5.4-2026-03-05_noneOpenAI | closed | effort none | 16.0%±3.7 | 30 Aug 2026 | Epoch AI log |
| 96 | qwen3.5-flash_noneAlibaba | not stated | effort none | 16.0%±3.7 | 27 Aug 2026 | Epoch AI log |
| 97 | qwen3.5-plusAlibaba | not stated | default | 16.0%±3.7 | 5 Aug 2026 | Epoch AI log |
| 98 | claude-opus-4-6Anthropic | closed | default | 15.0%±3.6 | 6 Aug 2026 | Epoch AI log |
| 99 | zai-org/GLM-5.2Epoch: glm-5.2_medium, Z.ai (Zhipu AI) | open | effort medium | 15.0%±3.6 | 27 Aug 2026 | Epoch AI log |
| 100 | gpt-5-2025-08-07_mediumOpenAI | closed | effort medium | 15.0%±3.6 | 27 Aug 2026 | Epoch AI log |
| 101 | gpt-5.1-2025-11-13_noneOpenAI | closed | effort none | 15.0%±3.6 | 27 Aug 2026 | Epoch AI log |
| 102 | qwen3.7-flashAlibaba | closed | default | 15.0%±3.6 | 28 Aug 2026 | Epoch AI log |
| 103 | muse-spark-1.3_xhighMeta AI | closed | effort xhigh | 14.1%±3.5 | 16 Sep 2026 | Epoch AI |
| 104 | claude-sonnet-4-6Anthropic | closed | default | 14.0%±3.5 | 6 Aug 2026 | Epoch AI log |
| 105 | gpt-5-2025-08-07_minimalOpenAI | closed | effort minimal | 14.0%±3.5 | 27 Aug 2026 | Epoch AI log |
| 106 | gpt-5.2-2025-12-11_noneOpenAI | closed | effort none | 14.0%±3.5 | 27 Aug 2026 | Epoch AI log |
| 107 | gpt-5.6-terra_noneOpenAI | closed | effort none | 14.0%±3.5 | 27 Aug 2026 | Epoch AI log |
| 108 | qwen3.7-flash_noneAlibaba | closed | effort none | 14.0%±3.5 | 27 Aug 2026 | Epoch AI log |
| 109 | claude-opus-4-7Anthropic | closed | default | 13.0%±3.4 | 6 Aug 2026 | Epoch AI log |
| 110 | gemini-3.5-flash-lite_minimalGoogle DeepMind | closed | effort minimal | 12.0%±3.3 | 5 Aug 2026 | Epoch AI log |
| 111 | gpt-4-0613OpenAI | closed | default | 12.0%±3.3 | 28 Aug 2026 | Epoch AI log |
| 112 | gpt-4o-mini-2024-07-18OpenAI | closed | default | 12.0%±3.3 | 27 Aug 2026 | Epoch AI log |
| 113 | gpt-5.6-luna_mediumOpenAI | closed | effort medium | 12.0%±3.3 | 27 Aug 2026 | Epoch AI log |
| 114 | moonshotai/Kimi-K2.6Epoch: kimi-k2.6_none, Moonshot | open | effort none | 12.0%±3.3 | 29 Aug 2026 | Epoch AI log |
| 115 | moonshotai/Kimi-K3Epoch: kimi-k3_low, Moonshot | open | effort low | 12.0%±3.3 | 29 Aug 2026 | Epoch AI log |
| 116 | qwen3.6-flashAlibaba | not stated | default | 12.0%±3.3 | 27 Aug 2026 | Epoch AI log |
| 117 | qwen3.6-plus_noneAlibaba | closed | effort none | 12.0%±3.3 | 27 Aug 2026 | Epoch AI log |
| 118 | gpt-5.4-mini-2026-03-17_noneOpenAI | closed | effort none | 11.0%±3.1 | 27 Aug 2026 | Epoch AI log |
| 119 | gpt-5-mini-2025-08-07_minimalOpenAI | closed | effort minimal | 10.0%±3.0 | 27 Aug 2026 | Epoch AI log |
| 120 | gpt-5.2-2025-12-11_lowOpenAI | closed | effort low | 10.0%±3.0 | 27 Aug 2026 | Epoch AI log |
| 121 | gpt-5.6-terra_mediumOpenAI | closed | effort medium | 10.0%±3.0 | 28 Aug 2026 | Epoch AI log |
| 122 | gpt-5-nano-2025-08-07_lowOpenAI | closed | effort low | 9.0%±2.9 | 27 Aug 2026 | Epoch AI log |
| 123 | gpt-5-nano-2025-08-07_mediumOpenAI | closed | effort medium | 9.0%±2.9 | 27 Aug 2026 | Epoch AI log |
| 124 | gpt-5.4-nano-2026-03-17_noneOpenAI | closed | effort none | 9.0%±2.9 | 27 Aug 2026 | Epoch AI log |
| 125 | Qwen/Qwen3-235B-A22B-Thinking-2507Epoch: Qwen3-235B-A22B-Thinking-2507, Alibaba | open | default | 9.0%±2.9 | 27 Aug 2026 | Epoch AI log |
| 126 | zai-org/GLM-5.3-FlashEpoch: glm-5.3-flash_max, Z.ai (Zhipu AI) | open | effort max | 8.0%±2.7 | 28 Aug 2026 | Epoch AI |
| 127 | gpt-5-nano-2025-08-07_highOpenAI | closed | effort high | 8.0%±2.7 | 27 Aug 2026 | Epoch AI log |
| 128 | gpt-5.4-mini-2026-03-17_lowOpenAI | closed | effort low | 8.0%±2.7 | 27 Aug 2026 | Epoch AI log |
| 129 | MiniMaxAI/MiniMax-M3Epoch: MiniMax-M3_none, MiniMax | open | effort none | 8.0%±2.7 | 27 Aug 2026 | Epoch AI log |
| 130 | claude-opus-4-6_lowAnthropic | closed | effort low | 7.0%±2.6 | 6 Aug 2026 | Epoch AI log |
| 131 | gpt-4.1-mini-2025-04-14OpenAI | closed | default | 7.0%±2.6 | 27 Aug 2026 | Epoch AI log |
| 132 | gpt-5.4-mini-2026-03-17_mediumOpenAI | closed | effort medium | 7.0%±2.6 | 27 Aug 2026 | Epoch AI log |
| 133 | gpt-6-luna_maxOpenAI | closed | effort max | 7.0%±2.6 | 22 Sep 2026 | Epoch AI |
| 134 | MiniMaxAI/MiniMax-M3Epoch: MiniMax-M3, MiniMax | open | default | 7.0%±2.6 | 27 Aug 2026 | Epoch AI log |
| 135 | o3-mini-2025-01-31_highOpenAI | closed | effort high | 7.0%±2.6 | 27 Aug 2026 | Epoch AI log |
| 136 | Qwen/Qwen3.6-27BEpoch: qwen3.6-27b_none, Alibaba | open | effort none | 7.0%±2.6 | 27 Aug 2026 | Epoch AI log |
| 137 | gpt-5.4-nano-2026-03-17_mediumOpenAI | closed | effort medium | 6.0%±2.4 | 27 Aug 2026 | Epoch AI log |
| 138 | thinkingmachines/Inkling-SmallEpoch: inkling-small_xhigh, Thinking Machines | open | effort xhigh | 6.0%±2.4 | 15 Aug 2026 | Epoch AI |
| 139 | gpt-5-mini-2025-08-07_highOpenAI | closed | effort high | 5.0%±2.2 | 27 Aug 2026 | Epoch AI log |
| 140 | gpt-5-nano-2025-08-07_minimalOpenAI | closed | effort minimal | 5.0%±2.2 | 27 Aug 2026 | Epoch AI log |
| 141 | gpt-5.4-nano-2026-03-17_highOpenAI | closed | effort high | 5.0%±2.2 | 27 Aug 2026 | Epoch AI log |
| 142 | o4-mini-2025-04-16_highOpenAI | closed | effort high | 5.0%±2.2 | 27 Aug 2026 | Epoch AI log |
| 143 | qwen3-max-2025-09-23Alibaba | closed | default | 5.0%±2.2 | 27 Aug 2026 | Epoch AI log |
| 144 | gpt-5-mini-2025-08-07_mediumOpenAI | closed | effort medium | 4.0%±2.0 | 27 Aug 2026 | Epoch AI log |
| 145 | gpt-3.5-turbo-0125OpenAI | closed | default | 3.0%±1.7 | 27 Aug 2026 | Epoch AI log |
| 146 | gpt-5.4-nano-2026-03-17_lowOpenAI | closed | effort low | 3.0%±1.7 | 27 Aug 2026 | Epoch AI log |
| 147 | openai/gpt-oss-120bEpoch: gpt-oss-120b_medium, OpenAI | open | effort medium | 2.0%±1.4 | 27 Aug 2026 | Epoch AI log |
| 148 | openai/gpt-oss-120bEpoch: gpt-oss-120b_high, OpenAI | open | effort high | 0.0% | 27 Aug 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