Pass IndexThe State of AISign in

Deepresearchbench — Epoch AI

Deepresearchbench — Epoch AI is run by Epoch AI. It has ranked 41 models, of which the catalogue holds 24, scoring from 0.351 to 0.553.

measured by Epoch AI · the board itself

PlaceModelMetricScore
1stof 41Claude Opus 4.6Average score (high)0.553
2ndof 41Claude Sonnet 4.6Average score (high)0.549
3rdof 41Claude Opus 4.5Average score (high)0.548
4thof 41GPT-5.5Average score (high)0.54
7thof 41Claude Sonnet 4.5Average score0.526
10thof 41Claude Opus 4.8Average score (high)0.502
11thof 41Gemini 3 FlashAverage score (low)0.498
12thof 41GPT-5Average score (low)0.496
19thof 41Claude Opus 4.1Average score0.483
22ndof 41Gemini 3.1 Pro PreviewAverage score (high)0.478
26thof 41Claude Opus 4Average score0.468
27thof 41Claude Sonnet 4Average score0.466
28thof 41gemini-3-pro-previewAverage score (low)0.463
29thof 41Claude Haiku 4.5Average score (low)0.455
30thof 41o3Average score (medium)0.452
32ndof 41Claude 3.7 SonnetAverage score0.436
33rdof 41Gemini 2.5 Pro Preview 06-05Average score0.428
34thof 41GPT-5.1Average score (low)0.428
35thof 41Gemini 2.5 ProAverage score0.415
36thof 41GPT-5.2Average score (low)0.411
37thof 41Gemini 3.1 Flash-LiteAverage score (low)0.373
39thof 41GPT-5.4 miniAverage score (low)0.363
40thof 41GPT-5.4Average score (low)0.351
41stof 41DeepSeek-R1 (0528)Average score0.351

Read from the board on 2026-09-16