Skip to content
Weights Not Walls
Menu

ProofBench

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.
Link
https://www.vals.ai/benchmarks/proof_bench
Score shown
Epoch's column "Accuracy", as a percentage, higher is better; random guessing scores about 0.0%. Error bars as the source gives them.
Results
85 in all: 30 for open-weight models, 55 for closed ones. 24 linked to a model page here.

Best open-weight result: moonshotai/Kimi-K3.1-Pro, 91.0%, number 10 of 85 results.

ProofBench results
#ModelWeightsSettingScoreDateSource
1claude-fable-5-1_maxAnthropicclosedeffort max100.0%not datedvals.ai
2claude-fable-5-2_maxAnthropicclosedeffort max100.0%not datedvals.ai
3claude-opus-5-5_maxAnthropicclosedeffort max100.0%not datedvals.ai
4claude-sonnet-5-5_maxAnthropicclosedeffort max100.0%not datedvals.ai
5claude-opus-5_maxAnthropicclosedeffort max99.0%±1.0not datedvals.ai
6gemini-4-argon_unknownGoogle DeepMindclosedeffort unknown99.0%not datedvals.ai
7gpt-6-astra_unknownOpenAIclosedeffort unknown99.0%not datedvals.ai
8gpt-6.1-sol_unknownOpenAIclosedeffort unknown99.0%not datedvals.ai
9claude-fable-5_maxAnthropicclosedeffort max; Reasoning effort: max95.0%±2.2not datedvals.ai
10moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro_unknown, Moonshotopeneffort unknown91.0%±2.9not datedvals.ai
11claude-haiku-5-5_maxAnthropicclosedeffort max89.0%not datedvals.ai
12moonshotai/Kimi-K3Epoch: kimi-k3_unknown, Moonshotopeneffort unknown87.0%±3.4not datedvals.ai
13muse-wonder-1.0_unknownMeta AIclosedeffort unknown84.0%not datedvals.ai
14gpt-5.6-sol_maxOpenAIclosedeffort max; Reasoning effort: max83.0%±3.8not datedvals.ai
15gpt-6-sol_unknownOpenAIclosedeffort unknown83.0%not datedvals.ai
16claude-sonnet-5_maxAnthropicclosedeffort max; Reasoning effort: max77.0%±4.2not datedvals.ai
17tencent/Hy4-previewEpoch: hy4-preview_unknown, Tencentopeneffort unknown75.0%not datedvals.ai
18gpt-5.6-terra_xhighOpenAIclosedeffort xhigh; Reasoning effort: xhigh74.0%±4.4not datedvals.ai
19mimo-v2.6-proXiaomi Corpno page matchopendefault70.0%not datedvals.ai
20claude-opus-4-8_maxAnthropicclosedeffort max; Reasoning effort: max69.0%±4.6not datedvals.ai
21gpt-6-luna_unknownOpenAIclosedeffort unknown64.0%not datedvals.ai
22mimo-v2.6-flashXiaomi Corpno page matchopendefault63.0%not datedvals.ai
23gpt-5.6-luna_maxOpenAIclosedeffort max; Reasoning effort: max60.0%±4.9not datedvals.ai
24gemini-3.7-flash_unknownGoogle DeepMindclosedeffort unknown58.0%not datedvals.ai
25muse-spark-1.3_maxMeta AIclosedeffort max58.0%not datedvals.ai
26qwen3.8-max_unknownAlibabaclosedeffort unknown58.0%not datedvals.ai
27deepseek-ai/DeepSeek-V4-Flash-0731Epoch: deepseek-v4-flash-0731_unknown, DeepSeekopeneffort unknown56.0%±5.0not datedvals.ai
28gpt-5.4-2026-03-05_xhighOpenAIclosedeffort xhigh; Reasoning effort: xhigh56.0%±5.0not datedvals.ai
29muse-spark-1.3_unknownMeta AIclosedeffort unknown55.0%not datedvals.ai
30claude-opus-4-7_maxAnthropicclosedeffort max; Reasoning effort: max54.0%±5.0not datedvals.ai
31deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_unknown, DeepSeekopeneffort unknown54.0%not datedvals.ai
32grok-4.6_unknownxAIclosedeffort unknown51.0%±5.0not datedvals.ai
33claude-opus-4-6_maxAnthropicclosedeffort max; Reasoning effort: max50.0%±5.0not datedvals.ai
34deepseek-ai/DeepSeek-V4-Pro-0813Epoch: deepseek-v4-pro-0813_unknown, DeepSeekopeneffort unknown50.0%±5.0not datedvals.ai
35gpt-5.5_xhighOpenAIclosedeffort xhigh; Reasoning effort: xhigh50.0%±5.0not datedvals.ai
36zai-org/GLM-5.3Epoch: glm-5.3_max, Z.ai (Zhipu AI)openeffort max49.0%not datedvals.ai
37gemini-3.8-flash_unknownGoogle DeepMindclosedeffort unknown48.0%not datedvals.ai
38claude-sonnet-4-6_maxAnthropicclosedeffort max; Reasoning effort: max45.0%±5.0not datedvals.ai
39muse-spark-1.2Meta AIcloseddefault43.0%not datedvals.ai
40step-5-preview_unknownStepFunclosedeffort unknown42.0%not datedvals.ai
41muse-spark-1.1Meta AIclosedReasoning effort: xhigh39.0%±4.9not datedvals.ai
42claude-opus-4-5-20251101_unknownAnthropicclosedeffort unknown; Reasoning effort: high36.0%±4.8not datedvals.ai
43gemini-3.6-flash_unknownGoogle DeepMindclosedeffort unknown; Reasoning effort: high36.0%±4.8not datedvals.ai
44zai-org/GLM-5.2Epoch: glm-5.2_max, Z.ai (Zhipu AI)openeffort max; Reasoning effort: max35.0%±4.8not datedvals.ai
45gemini-3.5-flash_highGoogle DeepMindclosedeffort high; Reasoning effort: high31.0%±4.6not datedvals.ai
46grok-4.5_highxAIclosedeffort high; Reasoning effort: high31.0%±4.6not datedvals.ai
47gemini-3.1-pro-previewGoogle DeepMindclosedReasoning effort: high26.0%±4.4not datedvals.ai
48grok-4.7_unknownxAIclosedeffort unknown26.0%not datedvals.ai
49qwen3.7-maxAlibabacloseddefault26.0%±4.4not datedvals.ai
50zai-org/GLM-5.1Epoch: glm-5.1, Z.ai (Zhipu AI)opendefault22.2%±4.2not datedvals.ai
51XiaomiMiMo/MiMo-V2.5-ProEpoch: mimo-v2.5-pro, Xiaomi Corpopendefault22.0%±4.2not datedvals.ai
52zai-org/GLM-5.3-FlashEpoch: glm-5.3-flash_max, Z.ai (Zhipu AI)openeffort max21.0%±4.1not datedvals.ai
53gpt-5.4-mini-2026-03-17_xhighOpenAIclosedeffort xhigh; Reasoning effort: xhigh21.0%±4.1not datedvals.ai
54gemini-3-pro-previewGoogle DeepMindclosedReasoning effort: high20.0%±4.0not datedvals.ai
55claude-sonnet-4-5-20250929_unknownAnthropicclosedeffort unknown19.0%±3.9not datedvals.ai
56beam-1.0_unknownReflectionno page matchopeneffort unknown18.0%±3.9not datedvals.ai
57gpt-5-2025-08-07_highOpenAIclosedeffort high; Reasoning effort: high18.0%±3.8not datedvals.ai
58MiniMaxAI/MiniMax-M3Epoch: MiniMax-M3, MiniMaxopendefault18.0%±3.9not datedvals.ai
59muse-sparkMeta AIcloseddefault17.0%±3.8not datedvals.ai
60deepseek-ai/DeepSeek-V4-ProEpoch: deepseek-v4-pro_max, DeepSeekopeneffort max; Reasoning effort: max16.0%±3.7not datedvals.ai
61moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshotopendefault16.0%±3.7not datedvals.ai
62XiaomiMiMo/MiMo-V2.5Epoch: mimo-v2.5, Xiaomi Corpopendefault16.0%±3.7not datedvals.ai
63Qwen/Qwen3.8-27BEpoch: qwen3.8-27b, Alibabaopendefault16.0%not datedvals.ai
64gemini-3-flash-previewGoogle DeepMindclosedReasoning effort: high15.0%±3.6not datedvals.ai
65gpt-5.2-2025-12-11_xhighOpenAIclosedeffort xhigh; Reasoning effort: xhigh15.0%±3.6not datedvals.ai
66grok-4.20-0309-reasoningxAIcloseddefault14.0%±3.5not datedvals.ai
67gemini-3.5-flash-liteGoogle DeepMindclosedReasoning effort: high13.0%±3.4not datedvals.ai
68gpt-5-nano-2025-08-07_highOpenAIclosedeffort high; Reasoning effort: high12.0%±3.3not datedvals.ai
69grok-4.3_highxAIclosedeffort high; Reasoning effort: high11.0%±3.1not datedvals.ai
70mistral-large-4-0_unknownMistral AIno page matchopeneffort unknown10.0%±3.0not datedvals.ai
71gpt-5-mini-2025-08-07_highOpenAIclosedeffort high; Reasoning effort: high9.0%±2.9not datedvals.ai
72gpt-5.1-codex-maxOpenAIclosedReasoning effort: high9.0%±2.9not datedvals.ai
73mistral-medium-2604Mistral AIno page matchopenReasoning effort: high9.0%±2.9not datedvals.ai
74deepseek-ai/DeepSeek-V3.2Epoch: fireworks/deepseek-v3p2, DeepSeekopenReasoning effort: none8.0%±2.7not datedvals.ai
75zai-org/GLM-4.7Epoch: glm-4.7, Z.ai (Zhipu AI)opendefault6.0%±2.4not datedvals.ai
76thinkingmachines/Inkling-SmallEpoch: inkling-small, Thinking MachinesopenReasoning effort: 0.996.0%±2.4not datedvals.ai
77gpt-5.4-nano-2026-03-17_highOpenAIclosedeffort high; Reasoning effort: high5.0%±2.2not datedvals.ai
78grok-4-1-fast-reasoningxAIcloseddefault4.0%±2.0not datedvals.ai
79MiniMaxAI/MiniMax-M2.5Epoch: MiniMax-M2.5, MiniMaxopendefault4.0%±2.0not datedvals.ai
80mercury-2.5_unknownInception Labsclosedeffort unknown3.0%not datedvals.ai
81MiniMaxAI/MiniMax-M2.7Epoch: MiniMax-M2.7, MiniMaxopendefault3.0%±1.7not datedvals.ai
82nemotron-3-ultraNvidiano page matchopendefault2.0%±1.4not datedvals.ai
83thinkingmachines/InklingEpoch: Inkling, Thinking Machinesopendefault0.0%not datedvals.ai
84poolside/Laguna-M.1Epoch: laguna-m.1, Poolsideopendefault0.0%not datedvals.ai
85poolside/Laguna-XS.2Epoch: laguna-xs.2, Poolsideopendefault0.0%not datedvals.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