Inference
EN
Epoch AI

Chess Puzzles

Evaluated by Epoch AI on their own harness.

ModelsModels with a result.90
Top resultBest result on this benchmark.79.0%Claude Fable 5.2
Top-3 spreadPoints from first to third.5.0 pts
YearYear of release.—

Result and price

80 of 90 results

Results

90 results
#ModelResultThe score from the source.
01Claude Fable 5.2max effort79.0% ±4.1
02Gemini 4 Carbonhigh effort78.0% ±4.2
03GPT-6 Astramax effort74.0% ±4.4
04GPT-6.1 Solmax effort68.0% ±4.7
05Claude Opus 5.5max effort67.0% ±4.7
06Gemini 3.7 Flashhigh effort66.0% ±4.8
07GPT-5.5xhigh effort63.0% ±4.9
08GPT-5.5 Proxhigh effort62.0% ±4.9
09GPT-5.2 Proxhigh effort60.0% ±4.9
10Gemini 4 Argonhigh effort59.0% ±4.9
10Claude Sonnet 5.5max effort59.0% ±4.9
10GPT-5.4 Proxhigh effort59.0% ±4.9
10GPT-5.6 Solmax effort59.0% ±4.9
14Gemini 3.8 Flashhigh effort58.0% ±5.0
15GPT-6 Solmax effort56.0% ±5.0
16Claude Haiku 5.5max effort55.0% ±5.0
17Claude Opus 4.8max effort52.0% ±5.0
18Gemini 3.1 Prohigh effort51.0% ±5.0
19Claude Fable 5.1max effort48.0% ±5.0
19DeepSeek V4.1 Flash 1017max effort48.0% ±5.0
19GPT-5.4xhigh effort48.0% ±5.0
19GPT-5.2xhigh effort48.0% ±5.0
19GPT-5 Prohigh effort48.0% ±5.0
24Claude Fable 5max effort47.0% ±5.0
24GPT-5.6 Terramax effort47.0% ±5.0
26Qwen3.8 Maxxhigh effort46.0% ±5.0
26DeepSeek V4 Pro 0813max effort46.0% ±5.0
28Claude Sonnet 5max effort45.0% ±5.0
28Gemini 3.5 Flashhigh effort45.0% ±5.0
28GPT-5.4 nanoxhigh effort45.0% ±5.0
31Claude Opus 5max effort43.0% ±5.0
31GPT-5.6 Lunamax effort43.0% ±5.0
33GPT-5.1high effort42.0% ±5.0
34Grok 4.7xhigh effort41.0% ±4.9
34Claude Opus 4.7max effort41.0% ±4.9
36Gemini 3.6 Flashhigh effort40.0% ±4.9
36Gemini 3 Flashhigh effort40.0% ±4.9
36Doubao Seed Evolving (Nov 2026)high effort40.0% ±4.9
39Muse Spark 1.3max effort37.0% ±4.9
39GLM-5.3max effort37.0% ±4.9
39Qwen3.8 Max (0902)xhigh effort37.0% ±4.9
42Kimi K3max effort36.0% ±4.8
42GPT-6 Lunamax effort36.0% ±4.8
42GPT-5high effort36.0% ±4.8
42o3high effort36.0% ±4.8
46Grok 4.5high effort35.0% ±4.8
47Muse Spark 1.1xhigh effort34.0% ±4.8
48Grok 4.6xhigh effort33.0% ±4.7
49DeepSeek V4 Flash 0731max effort32.0% ±4.7
50Mistral Large 4high effort31.0% ±4.7
51Grok 4.2029.0% ±4.6
52Muse Spark 1.2xhigh effort27.0% ±4.5
52Qwen3.7 Max27.0% ±4.5
54Doubao Seed 2.1 Turbohigh effort26.0% ±4.4
55GPT-5 minihigh effort24.0% ±4.3
56Claude Sonnet 4.6max effort23.0% ±4.2
57GLM-5.2max effort22.0% ±4.2
57GPT-5 nanohigh effort22.0% ±4.2
59GLM-5.3-Flashmax effort21.0% ±4.1
59Claude Opus 4.6max effort21.0% ±4.1
59Inklingxhigh effort21.0% ±4.1
59GLM-4.721.0% ±4.1
59Doubao Seed 2.1 Prohigh effort21.0% ±4.1
64Kimi K2.620.0% ±4.0
64GLM-5.120.0% ±4.0
64Doubao Seed 2.0 Lite (Apr 2026)high effort20.0% ±4.0
64Doubao-Seed-2.0 Prohigh effort20.0% ±4.0
68DeepSeek V4 Pro 0423max effort19.0% ±3.9
69Grok 4.3 Betahigh effort18.0% ±3.9
70GPT-5.4 minixhigh effort17.0% ±3.8
71Qwen3.6 Plus16.0% ±3.7
71Gemini 3.5 Flash-Litehigh effort16.0% ±3.7
73o3-minihigh effort15.0% ±3.6
74Kimi K2.7 Code14.0% ±3.5
74Qwen 3.5 Flash (hosted 35B-A3B)14.0% ±3.5
76o4-minihigh effort13.0% ±3.4
77Qwen3.6 Flash12.0% ±3.3
78GPT-4.111.0% ±3.1
79Claude Sonnet 4.510.0% ±3.0
79o1high effort10.0% ±3.0
81GPT-4o9.0% ±2.9
81Qwen 3.5 Plus (hosted 397B-A17B)9.0% ±2.9
83MiniMax M3none effort8.0% ±2.7
84Claude Haiku 4.56.5% ±2.7
85Claude Opus 4.55.0% ±2.2
85Doubao Seed 1.65.0% ±2.2
87Claude 3 Opus4.0% ±2.0
88GPT-4o mini3.0% ±1.7
89GPT-4.1 mini1.0% ±1.0
90GPT-4.1 nano0.0%