Skip to content
Weights Not Walls
Menu

APEX-Agents

Reported

No 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.

APEX-Agents results
#ModelWeightsSettingScoreDateSource
1gemini-4-carbon_unknownGoogle DeepMindclosedeffort unknown84.2%error bar left outnot datedvia Epoch AI
2gemini-4-argon_unknownGoogle DeepMindclosedeffort unknown82.4%error bar left outnot datedvia Epoch AI
3claude-fable-5-2_unknownAnthropicclosedeffort unknown80.1%error bar left outnot datedvia Epoch AI
4gemini-4-hydrogen_unknownGoogle DeepMindclosedeffort unknown78.0%error bar left outnot datedvia Epoch AI
5claude-sonnet-5-5_maxAnthropicclosedeffort max75.5%error bar left outnot datedvia Epoch AI
6claude-opus-5-5_maxAnthropicclosedeffort max73.5%error bar left outnot datedvia Epoch AI
7grok-4.7_unknownxAIclosedeffort unknown70.0%error bar left outnot datedvia Epoch AI
8deepseek-ai/DeepSeek-V4.1-Flash-1017Epoch: deepseek-v4.1-flash-1017_max, DeepSeekopeneffort max69.8%error bar left outnot datedvia Epoch AI
9claude-fable-5-1_unknownAnthropicclosedeffort unknown68.6%error bar left outnot datedvia Epoch AI
10gpt-6.1-terra_maxOpenAIclosedeffort max68.3%error bar left outnot datedvia Epoch AI
11gemini-3.7-flash_unknownGoogle DeepMindclosedeffort unknown67.8%error bar left outnot datedvia Epoch AI
12muse-wonder-1.0_maxMeta AIclosedeffort max66.1%error bar left outnot datedvia Epoch AI
13claude-opus-5_maxAnthropicclosedeffort max65.8%error bar left outnot datedvia Epoch AI
14moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro_unknown, Moonshotopeneffort unknown65.5%error bar left outnot datedvia Epoch AI
15grok-4.6_unknownxAIclosedeffort unknown65.3%error bar left outnot datedvia Epoch AI
16gpt-6-astra_unknownOpenAIclosedeffort unknown64.7%error bar left outnot datedvia Epoch AI
17gemini-3.8-flash_unknownGoogle DeepMindclosedeffort unknown64.3%error bar left outnot datedvia Epoch AI
18muse-ember-1.0_unknownMeta AIclosedeffort unknown64.3%error bar left outnot datedvia Epoch AI
19claude-fable-5Anthropiccloseddefault63.6%error bar left outnot datedvia Epoch AI
20qwen3.8-max_unknownAlibabaclosedeffort unknown63.3%error bar left outnot datedvia Epoch AI
21gpt-6.1-sol_maxOpenAIclosedeffort max60.0%error bar left outnot datedvia Epoch AI
22claude-fable-5-1_highAnthropicclosedeffort high59.7%error bar left outnot datedvia Epoch AI
23muse-spark-1.3_unknownMeta AIclosedeffort unknown58.6%error bar left outnot datedvia Epoch AI
24gpt-5.6-terra_maxOpenAIclosedeffort max58.2%error bar left outnot datedvia Epoch AI
25zai-org/GLM-5.3Epoch: glm-5.3_unknown, Z.ai (Zhipu AI)openeffort unknown56.6%error bar left outnot datedvia Epoch AI
26grok-4.5_unknownxAIclosedeffort unknown56.2%error bar left outnot datedvia Epoch AI
27claude-haiku-5-5_maxAnthropicclosedeffort max56.1%error bar left outnot datedvia Epoch AI
28gpt-5.5_unknownOpenAIclosedeffort unknown55.1%error bar left outnot datedvia Epoch AI
29claude-sonnet-5_unknownAnthropicclosedeffort unknown54.5%error bar left outnot datedvia Epoch AI
30mimo-v2.6-proXiaomi Corpno page matchopendefault54.4%error bar left outnot datedvia Epoch AI
31gpt-6-sol_maxOpenAIclosedeffort max54.3%error bar left outnot datedvia Epoch AI
32zai-org/GLM-5.3-FlashEpoch: glm-5.3-flash_unknown, Z.ai (Zhipu AI)openeffort unknown52.8%error bar left outnot datedvia Epoch AI
33claude-opus-5-5_mediumAnthropicclosedeffort medium52.5%error bar left outnot datedvia Epoch AI
34gpt-5.4-2026-03-05_unknownOpenAIclosedeffort unknown52.4%error bar left outnot datedvia Epoch AI
35mistral-large-4-0_highMistral AIno page matchopeneffort high51.6%error bar left outnot datedvia Epoch AI
36gpt-5.6-sol_promaxOpenAIcloseddefault51.4%error bar left outnot datedvia Epoch AI
37claude-opus-4-7_maxAnthropicclosedeffort max49.2%error bar left outnot datedvia Epoch AI
38claude-opus-4-8_maxAnthropicclosedeffort max48.9%error bar left outnot datedvia Epoch AI
39beam-1.0_maxReflectionno page matchopeneffort max48.8%error bar left outnot datedvia Epoch AI
40moonshotai/Kimi-K3Epoch: kimi-k3_unknown, Moonshotopeneffort unknown47.8%error bar left outnot datedvia Epoch AI
41muse-spark-1.3_maxMeta AIclosedeffort max47.6%error bar left outnot datedvia Epoch AI
42Qwen/Qwen3.8-27BEpoch: qwen3.8-27b, Alibabaopendefault47.5%error bar left outnot datedvia Epoch AI
43deepseek-ai/DeepSeek-V4-Pro-0813Epoch: deepseek-v4-pro-0813_unknown, DeepSeekopeneffort unknown47.3%error bar left outnot datedvia Epoch AI
44gemini-3.6-flash_unknownGoogle DeepMindclosedeffort unknown46.9%error bar left outnot datedvia Epoch AI
45thinkingmachines/Inkling-2-SmallEpoch: inkling-2-small_unknown, Thinking Machinesopeneffort unknown46.8%error bar left outnot datedvia Epoch AI
46claude-opus-4-6_maxAnthropicclosedeffort max46.3%error bar left outnot datedvia Epoch AI
47muse-spark-1.2Meta AIcloseddefault45.4%error bar left outnot datedvia Epoch AI
48zai-org/GLM-5.2Epoch: glm-5.2_unknown, Z.ai (Zhipu AI)openeffort unknown45.2%error bar left outnot datedvia Epoch AI
49claude-sonnet-5-5_mediumAnthropicclosedeffort medium44.6%error bar left outnot datedvia Epoch AI
50gpt-6-luna_maxOpenAIclosedeffort max44.3%error bar left outnot datedvia Epoch AI
51claude-sonnet-4-6_highAnthropicclosedeffort high43.0%error bar left outnot datedvia Epoch AI
52gpt-5.6-luna_unknownOpenAIclosedeffort unknown43.0%error bar left outnot datedvia Epoch AI
53zai-org/GLM-5.1Epoch: glm-5.1, Z.ai (Zhipu AI)opendefault40.9%error bar left outnot datedvia Epoch AI
54deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_unknown, DeepSeekopeneffort unknown39.5%error bar left outnot datedvia Epoch AI
55MiniMaxAI/MiniMax-M3Epoch: MiniMax-M3, MiniMaxopendefault37.7%error bar left outnot datedvia Epoch AI
56moonshotai/Kimi-K2.7-CodeEpoch: kimi-k2.7-code, Moonshotopendefault37.6%error bar left outnot datedvia Epoch AI
57gemini-3.1-pro-previewGoogle DeepMindcloseddefault35.1%error bar left outnot datedvia Epoch AI
58thinkingmachines/InklingEpoch: Inkling, Thinking Machinesopendefault33.8%error bar left outnot datedvia Epoch AI
59meta-models/Muse-Spark-Open-816B-A42BEpoch: Muse-Spark-Open-816B-A42B_unknown, Meta AIopeneffort unknown33.4%error bar left outnot datedvia Epoch AI
60muse-spark-1.1Meta AIcloseddefault31.8%error bar left outnot datedvia Epoch AI
61gemini-3.5-flash-liteGoogle DeepMindcloseddefault29.3%error bar left outnot datedvia Epoch AI
62gemini-3.5-flash_unknownGoogle DeepMindclosedeffort unknown27.5%error bar left outnot datedvia Epoch AI
63Qwen/Qwen3.5-397B-A17BEpoch: qwen3.5-397b-a17b, Alibabaopendefault24.9%error bar left outnot datedvia Epoch AI
64nemotron-3-ultraNvidiano page matchopendefault22.7%error bar left outnot datedvia Epoch AI
65deepseek-ai/DeepSeek-V3.2Epoch: deepseek/deepseek-v3.2, DeepSeekopendefault21.3%error bar left outnot datedvia Epoch AI
66openai/gpt-oss-120bEpoch: gpt-oss-120b, OpenAIopendefault4.4%error bar left outnot datedvia 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