APEX-Agents
ReportedNo description yet: we only describe an eval set when we can do it correctly.
- Kind
- Published by a third party's leaderboard, collected by Epoch AI. The leaderboard is named and linked.
- Score shown
- Epoch's column "Pass@1 score", as a percentage, higher is better; random guessing scores about 0.0%. Error bars as the source gives them.
- Results
- 66 in all: 22 for open-weight models, 44 for closed ones. 18 linked to a model page here.
Best open-weight result: deepseek-ai/DeepSeek-V4.1-Flash-1017, 69.8%, number 8 of 66 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | gemini-4-carbon_unknownGoogle DeepMind | closed | effort unknown | 84.2%error bar left out | not dated | via Epoch AI |
| 2 | gemini-4-argon_unknownGoogle DeepMind | closed | effort unknown | 82.4%error bar left out | not dated | via Epoch AI |
| 3 | claude-fable-5-2_unknownAnthropic | closed | effort unknown | 80.1%error bar left out | not dated | via Epoch AI |
| 4 | gemini-4-hydrogen_unknownGoogle DeepMind | closed | effort unknown | 78.0%error bar left out | not dated | via Epoch AI |
| 5 | claude-sonnet-5-5_maxAnthropic | closed | effort max | 75.5%error bar left out | not dated | via Epoch AI |
| 6 | claude-opus-5-5_maxAnthropic | closed | effort max | 73.5%error bar left out | not dated | via Epoch AI |
| 7 | grok-4.7_unknownxAI | closed | effort unknown | 70.0%error bar left out | not dated | via Epoch AI |
| 8 | deepseek-ai/DeepSeek-V4.1-Flash-1017Epoch: deepseek-v4.1-flash-1017_max, DeepSeek | open | effort max | 69.8%error bar left out | not dated | via Epoch AI |
| 9 | claude-fable-5-1_unknownAnthropic | closed | effort unknown | 68.6%error bar left out | not dated | via Epoch AI |
| 10 | gpt-6.1-terra_maxOpenAI | closed | effort max | 68.3%error bar left out | not dated | via Epoch AI |
| 11 | gemini-3.7-flash_unknownGoogle DeepMind | closed | effort unknown | 67.8%error bar left out | not dated | via Epoch AI |
| 12 | muse-wonder-1.0_maxMeta AI | closed | effort max | 66.1%error bar left out | not dated | via Epoch AI |
| 13 | claude-opus-5_maxAnthropic | closed | effort max | 65.8%error bar left out | not dated | via Epoch AI |
| 14 | moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro_unknown, Moonshot | open | effort unknown | 65.5%error bar left out | not dated | via Epoch AI |
| 15 | grok-4.6_unknownxAI | closed | effort unknown | 65.3%error bar left out | not dated | via Epoch AI |
| 16 | gpt-6-astra_unknownOpenAI | closed | effort unknown | 64.7%error bar left out | not dated | via Epoch AI |
| 17 | gemini-3.8-flash_unknownGoogle DeepMind | closed | effort unknown | 64.3%error bar left out | not dated | via Epoch AI |
| 18 | muse-ember-1.0_unknownMeta AI | closed | effort unknown | 64.3%error bar left out | not dated | via Epoch AI |
| 19 | claude-fable-5Anthropic | closed | default | 63.6%error bar left out | not dated | via Epoch AI |
| 20 | qwen3.8-max_unknownAlibaba | closed | effort unknown | 63.3%error bar left out | not dated | via Epoch AI |
| 21 | gpt-6.1-sol_maxOpenAI | closed | effort max | 60.0%error bar left out | not dated | via Epoch AI |
| 22 | claude-fable-5-1_highAnthropic | closed | effort high | 59.7%error bar left out | not dated | via Epoch AI |
| 23 | muse-spark-1.3_unknownMeta AI | closed | effort unknown | 58.6%error bar left out | not dated | via Epoch AI |
| 24 | gpt-5.6-terra_maxOpenAI | closed | effort max | 58.2%error bar left out | not dated | via Epoch AI |
| 25 | zai-org/GLM-5.3Epoch: glm-5.3_unknown, Z.ai (Zhipu AI) | open | effort unknown | 56.6%error bar left out | not dated | via Epoch AI |
| 26 | grok-4.5_unknownxAI | closed | effort unknown | 56.2%error bar left out | not dated | via Epoch AI |
| 27 | claude-haiku-5-5_maxAnthropic | closed | effort max | 56.1%error bar left out | not dated | via Epoch AI |
| 28 | gpt-5.5_unknownOpenAI | closed | effort unknown | 55.1%error bar left out | not dated | via Epoch AI |
| 29 | claude-sonnet-5_unknownAnthropic | closed | effort unknown | 54.5%error bar left out | not dated | via Epoch AI |
| 30 | mimo-v2.6-proXiaomi Corpno page match | open | default | 54.4%error bar left out | not dated | via Epoch AI |
| 31 | gpt-6-sol_maxOpenAI | closed | effort max | 54.3%error bar left out | not dated | via Epoch AI |
| 32 | zai-org/GLM-5.3-FlashEpoch: glm-5.3-flash_unknown, Z.ai (Zhipu AI) | open | effort unknown | 52.8%error bar left out | not dated | via Epoch AI |
| 33 | claude-opus-5-5_mediumAnthropic | closed | effort medium | 52.5%error bar left out | not dated | via Epoch AI |
| 34 | gpt-5.4-2026-03-05_unknownOpenAI | closed | effort unknown | 52.4%error bar left out | not dated | via Epoch AI |
| 35 | mistral-large-4-0_highMistral AIno page match | open | effort high | 51.6%error bar left out | not dated | via Epoch AI |
| 36 | gpt-5.6-sol_promaxOpenAI | closed | default | 51.4%error bar left out | not dated | via Epoch AI |
| 37 | claude-opus-4-7_maxAnthropic | closed | effort max | 49.2%error bar left out | not dated | via Epoch AI |
| 38 | claude-opus-4-8_maxAnthropic | closed | effort max | 48.9%error bar left out | not dated | via Epoch AI |
| 39 | beam-1.0_maxReflectionno page match | open | effort max | 48.8%error bar left out | not dated | via Epoch AI |
| 40 | moonshotai/Kimi-K3Epoch: kimi-k3_unknown, Moonshot | open | effort unknown | 47.8%error bar left out | not dated | via Epoch AI |
| 41 | muse-spark-1.3_maxMeta AI | closed | effort max | 47.6%error bar left out | not dated | via Epoch AI |
| 42 | Qwen/Qwen3.8-27BEpoch: qwen3.8-27b, Alibaba | open | default | 47.5%error bar left out | not dated | via Epoch AI |
| 43 | deepseek-ai/DeepSeek-V4-Pro-0813Epoch: deepseek-v4-pro-0813_unknown, DeepSeek | open | effort unknown | 47.3%error bar left out | not dated | via Epoch AI |
| 44 | gemini-3.6-flash_unknownGoogle DeepMind | closed | effort unknown | 46.9%error bar left out | not dated | via Epoch AI |
| 45 | thinkingmachines/Inkling-2-SmallEpoch: inkling-2-small_unknown, Thinking Machines | open | effort unknown | 46.8%error bar left out | not dated | via Epoch AI |
| 46 | claude-opus-4-6_maxAnthropic | closed | effort max | 46.3%error bar left out | not dated | via Epoch AI |
| 47 | muse-spark-1.2Meta AI | closed | default | 45.4%error bar left out | not dated | via Epoch AI |
| 48 | zai-org/GLM-5.2Epoch: glm-5.2_unknown, Z.ai (Zhipu AI) | open | effort unknown | 45.2%error bar left out | not dated | via Epoch AI |
| 49 | claude-sonnet-5-5_mediumAnthropic | closed | effort medium | 44.6%error bar left out | not dated | via Epoch AI |
| 50 | gpt-6-luna_maxOpenAI | closed | effort max | 44.3%error bar left out | not dated | via Epoch AI |
| 51 | claude-sonnet-4-6_highAnthropic | closed | effort high | 43.0%error bar left out | not dated | via Epoch AI |
| 52 | gpt-5.6-luna_unknownOpenAI | closed | effort unknown | 43.0%error bar left out | not dated | via Epoch AI |
| 53 | zai-org/GLM-5.1Epoch: glm-5.1, Z.ai (Zhipu AI) | open | default | 40.9%error bar left out | not dated | via Epoch AI |
| 54 | deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_unknown, DeepSeek | open | effort unknown | 39.5%error bar left out | not dated | via Epoch AI |
| 55 | MiniMaxAI/MiniMax-M3Epoch: MiniMax-M3, MiniMax | open | default | 37.7%error bar left out | not dated | via Epoch AI |
| 56 | moonshotai/Kimi-K2.7-CodeEpoch: kimi-k2.7-code, Moonshot | open | default | 37.6%error bar left out | not dated | via Epoch AI |
| 57 | gemini-3.1-pro-previewGoogle DeepMind | closed | default | 35.1%error bar left out | not dated | via Epoch AI |
| 58 | thinkingmachines/InklingEpoch: Inkling, Thinking Machines | open | default | 33.8%error bar left out | not dated | via Epoch AI |
| 59 | meta-models/Muse-Spark-Open-816B-A42BEpoch: Muse-Spark-Open-816B-A42B_unknown, Meta AI | open | effort unknown | 33.4%error bar left out | not dated | via Epoch AI |
| 60 | muse-spark-1.1Meta AI | closed | default | 31.8%error bar left out | not dated | via Epoch AI |
| 61 | gemini-3.5-flash-liteGoogle DeepMind | closed | default | 29.3%error bar left out | not dated | via Epoch AI |
| 62 | gemini-3.5-flash_unknownGoogle DeepMind | closed | effort unknown | 27.5%error bar left out | not dated | via Epoch AI |
| 63 | Qwen/Qwen3.5-397B-A17BEpoch: qwen3.5-397b-a17b, Alibaba | open | default | 24.9%error bar left out | not dated | via Epoch AI |
| 64 | nemotron-3-ultraNvidiano page match | open | default | 22.7%error bar left out | not dated | via Epoch AI |
| 65 | deepseek-ai/DeepSeek-V3.2Epoch: deepseek/deepseek-v3.2, DeepSeek | open | default | 21.3%error bar left out | not dated | via Epoch AI |
| 66 | openai/gpt-oss-120bEpoch: gpt-oss-120b, OpenAI | open | default | 4.4%error bar left out | not dated | via Epoch AI |
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