DeepResearch Bench
Rank #36 of 51 · 22 models scored · top score 55.31
unknown
Strongest on harness comparability, weakest on contamination resistance. Caveat: the public test set leaves contamination on the table.
Weighted composite. Each signal is computed from the source-backed leaderboard below.
Does it still separate models?
How far from ceiling / clustered at the top?
Public vs held-out; training-leak risk.
Apples-to-apples, or mixed / vendor-optimized?
How old is the benchmark?
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | Claude Opus 4.6 | 55.31 | unverified | 2026-02-05 |
| 2 | GPT-5 | 55.13 | unverified | 2025-08-07 |
| 3 | Claude Sonnet 4.6 | 54.87 | unverified | 2026-02-17 |
| 4 | GPT-5.5 | 54.01 | unverified | 2026-04-23 |
| 5 | Claude Sonnet 4.5 | 52.6 | unverified | 2025-09-29 |
| 6 | Claude Opus 4.8 | 50.23 | unverified | 2026-05-28 |
| 7 | Gemini 2.5 Pro (Jun 2025) | 49.71 | unverified | 2025-06-05 |
| 8 | Claude Opus 4.1 | 49.7 | unverified | 2025-08-05 |
| 9 | Claude Opus 4 | 49 | unverified | 2025-05-22 |
| 10 | Grok 4 | 47.9 | unverified | 2025-07-09 |
| 11 | Claude Sonnet 4 | 47.8 | unverified | 2025-05-22 |
| 12 | o3 | 46.6 | unverified | 2024-12-20 |
| 13 | Claude 3.7 Sonnet | 43.6 | unverified | 2025-02-24 |
| 14 | GPT-5.1 | 42.79 | unverified | 2025-11-13 |
| 15 | GPT-5.2 | 41.12 | unverified | 2025-12-11 |
| 16 | Gemini 3.1 Flash-Lite | 36.39 | unverified | 2026-03-03 |
| 17 | GPT-5.4 Mini | 36.26 | unverified | 2026-03-17 |
| 18 | DeepSeek-R1 (May 2025) | 35.1 | unverified | 2025-05-28 |
| 19 | GPT-5.4 | 35.09 | unverified | 2026-03-05 |
| 20 | Gemini 2.5 Pro (May 2025) | 31.9 | unverified | 2025-05-06 |
| 21 | GPT-4.1 | 29.31 | unverified | 2025-04-14 |
| 22 | Gemini 2.5 Flash (Apr 2025) | 29.19 | unverified | 2025-04-17 |