Pass IndexThe State of AISign in

Dtbench — Epoch AI

Dtbench — Epoch AI is run by Epoch AI. It has ranked 161 models, of which the catalogue holds 136, scoring from 41.65 to 98.4.

measured by Epoch AI · the board itself

PlaceModelMetricScore
1stof 161Claude Fable 5Accuracy (max)98.4
2ndof 161Claude Opus 5Accuracy (max)97.6
3rdof 161Grok 4.6Accuracy (xhigh)97.33
4thof 161Gemini 3.7 FlashAccuracy (high)96.8
5thof 161Grok 4.5Accuracy (high)96.53
7thof 161GPT-5.5Accuracy (xhigh)96
8thof 161GPT-5.5 ProAccuracy (xhigh)96
9thof 161GPT-5.6 SolAccuracy (max)95.47
10thof 161Gemini 3.1 Pro PreviewAccuracy (high)95.47
11thof 161Gemini 3.6 FlashAccuracy (high)95.47
12thof 161Claude Opus 4.8Accuracy (max)94.93
13thof 161Muse Spark 1.2Accuracy (xhigh)94.67
14thof 161Gemini 3.5 FlashAccuracy (high)94.67
15thof 161Claude Opus 4.7Accuracy (max)94.65
16thof 161GPT-5.4Accuracy (xhigh)94.4
17thof 161Muse Spark 1.1Accuracy (high)94.4
18thof 161GLM 5.2Accuracy (max)93.6
19thof 161GPT-5.6 TerraAccuracy (max)92.53
20thof 161Claude Sonnet 5Accuracy (max)92.48
21stof 161Qwen3.7 MaxAccuracy (max)92.27
22ndof 161Qwen 3.8 MaxAccuracy (xhigh)92
23rdof 161Claude Opus 4.6Accuracy (max)91.2
24thof 161Kimi K3Accuracy (max)91.2
25thof 161GPT-5.2Accuracy (xhigh)90.93
26thof 161DeepSeek V4 Flash (0731)Accuracy (max)90.93
27thof 161Kimi K2.6Accuracy90.93
28thof 161DeepSeek V4 ProAccuracy (max)90.67
29thof 161GPT-5Accuracy (high)90.67
30thof 161Grok 4.3Accuracy (high)90.67
31stof 161GPT-5.1Accuracy (high)90.13
32ndof 161Grok 4.20Accuracy90.13
33rdof 161Nemotron 3 UltraAccuracy90.13
34thof 161Claude Sonnet 4.6Accuracy (max)89.87
35thof 161Claude Opus 4.5Accuracy (high)89.87
36thof 161GPT-5.6 LunaAccuracy (max)89.07
37thof 161Gemini 3 FlashAccuracy (high)89.07
38thof 161Grok 4.1 Fast ReasoningAccuracy87.73
40thof 161InklingAccuracy (xhigh)87.47
41stof 161Qwen3.5 397B A17BAccuracy87.47
42ndof 161Qwen3.6 Max PreviewAccuracy (max)87.2
43rdof 161o3-proAccuracy (high)86.93
44thof 161DeepSeek V4 FlashAccuracy (max)86.4
45thof 161DeepSeek v3.2Accuracy85.6
46thof 161o3Accuracy (high)84.8
47thof 161MiMo-V2.5-ProAccuracy84.53
48thof 161Qwen3.5-122B-A10BAccuracy (none)84.27
49thof 161Qwen 3.7 PlusAccuracy84
50thof 161Gemini 3.5 Flash-LiteAccuracy (high)83.47
51stof 161Claude Sonnet 4.5Accuracy83.2
52ndof 161Qwen3.5-FlashAccuracy82.93
53rdof 161Grok 4 FastAccuracy82.67
54thof 161Gemma 4 31BAccuracy82.67
56thof 161Gemini 2.5 ProAccuracy82.4
57thof 161Qwen3.5-27BAccuracy82.4
59thof 161Qwen3.6 PlusAccuracy81.87
60thof 161Claude Opus 4Accuracy81.6
61stof 161DeepSeek V3.1 TerminusAccuracy81.33
62ndof 161Qwen-PlusAccuracy81.07
63rdof 161GPT-5 miniAccuracy (high)80.53
64thof 161Qwen 3.5 PlusAccuracy80.53
65thof 161GPT-5.4 nanoAccuracy (xhigh)80.27
66thof 161Qwen3 235B A22B Thinking 2507Accuracy80.27
67thof 161GPT-5.4 miniAccuracy (xhigh)80
68thof 161Claude Opus 4.1Accuracy80
69thof 161Qwen3.5-35B-A3BAccuracy80
70thof 161MiniMax M3Accuracy78.93
71stof 161Qwen3-235B-A22B-Instruct-2507Accuracy78.4
72ndof 161Qwen3.6 27BAccuracy78.13
73rdof 161o4-miniAccuracy (high)77.6
74thof 161Qwen3.6-FlashAccuracy77.07
75thof 161Claude Sonnet 4Accuracy77.07
76thof 161Gemini 3.1 Flash-LiteAccuracy (high)76.8
77thof 161Gemini 2.5 FlashAccuracy76.53
78thof 161GPT OSS 120BAccuracy (high)76.27
79thof 161Qwen3 235B A22BAccuracy75.73
81stof 161Gemma 4 26B A4BAccuracy74.93
82ndof 161o1Accuracy (high)74.67
83rdof 161Qwen3.6 35B A3BAccuracy73.87
84thof 161Claude Haiku 4.5Accuracy73.6
85thof 161Qwen3.5 9BAccuracy71.2
86thof 161Mistral Small 4Accuracy70.93
87thof 161Qwen3 30B A3B Thinking 2507Accuracy69.33
88thof 161o3-miniAccuracy (high)68.8
89thof 161GPT-4.1 miniAccuracy68.8
90thof 161GPT-4.1 fine-tunedAccuracy68.27
91stof 161GPT OSS 20BAccuracy (high)68
92ndof 161Claude 3.5 SonnetAccuracy67.84
94thof 161Qwen3 32BAccuracy67.47
96thof 161Qwen3 30B A3B Instruct 2507Accuracy67.2
98thof 161Mistral Large 3Accuracy65.07
99thof 161DeepSeek-V3 0324Accuracy64.8
100thof 161GPT-4o (2024-05-13)Accuracy64.53
101stof 161Qwen3 14BAccuracy64
102ndof 161gemini-2.0-flash-001Accuracy63.2
103rdof 161Qwen2.5 72B InstructAccuracy62.93
105thof 161GPT-4 (0613)Accuracy62.67
106thof 161deepseek-chatAccuracy62.67
107thof 161GPT-5 nanoAccuracy (high)62.67
108thof 161Mistral Medium 2505Accuracy62.27
109thof 161Llama-4-Maverick-17B-128E-InstructAccuracy61.87
110thof 161Claude 3 OpusAccuracy61.6
111thof 161GPT-4 TurboAccuracy61.6
112thof 161Meta-Llama-3.1-405B-InstructAccuracy61.38
113thof 161c4ai-command-a-03-2025Accuracy61.33
115thof 161Mistral Large 2407Accuracy61.21
117thof 161Qwen3 30B A3BAccuracy60.27
118thof 161Meta-Llama-3.1-70B-InstructAccuracy60
120thof 161Qwen3 8BAccuracy59.73
121stof 161Llama 3.3 70B InstructAccuracy59.47
123rdof 161mistral-small-2503Accuracy58.64
125thof 161Claude Haiku 3.5Accuracy56.66
127thof 161Mistral Large (24.02)Accuracy55.9
128thof 161Mixtral-8x22B-Instruct-v0.1Accuracy55.14
129thof 161c4ai-command-r-plus-08-2024Accuracy54.93
130thof 161GPT-4o-mini (2024-07-18)Accuracy54.4
131stof 161Meta-Llama-3-70B-InstructAccuracy54.16
133rdof 161Claude 3 SonnetAccuracy53.6
135thof 161Mistral Small (24.02)Accuracy52.96
136thof 161Gemini 2.0 Flash-LiteAccuracy52.53
137thof 161Gemma 3 27BAccuracy52.53
138thof 161GPT-4.1 nanoAccuracy52.53
139thof 161Claude 2.0Accuracy51.85
142ndof 161Claude 2.1Accuracy50.95
143rdof 161Llama 3.1 8B InstructAccuracy50.93
144thof 161Gemma 3 4BAccuracy50.93
145thof 161Claude 3 HaikuAccuracy50.13
149thof 161Gemma 3 12BAccuracy48.8
150thof 161Mistral NemoAccuracy48.61
151stof 161GPT-3.5 TurboAccuracy48.53
152ndof 161Gemma 2 27BAccuracy48
153rdof 161Qwen2.5 7B InstructAccuracy47.73
154thof 161c4ai-command-r-08-2024Accuracy46.4
158thof 161Llama 3 8B InstructAccuracy43.93
159thof 161open-mistral-7bAccuracy42.54
160thof 161Llama-2-13b-chatAccuracy42.24
161stof 161Llama-2-70b-chatAccuracy41.65

Read from the board on 2026-09-16