ResearchCodeBench — leaderboard
Metric: Task Success Rate (%). Source: researchcodebench.github.io. 32 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 05-06) | 64.2 |
| 2 | O3 (High) | 59.4 |
| 3 | Gemini 2.5 Pro (Preview 03-25) | 59 |
| 4 | O4 Mini (High) | 58.5 |
| 5 | O3 Mini (High) | 52.4 |
| 6 | Claude 3.7 Sonnet | 51.9 |
| 7 | GPT-4.1 | 50 |
| 8 | Claude 3.5 Sonnet | 48.6 |
| 9 | O1 (High) | 48.1 |
| 10 | DeepSeek R1 | 45.8 |
| 11 | Gemini 2.5 Flash (Preview 04-17) | 45.3 |
| 12 | Grok 3 | 42.9 |
| 13 | GPT-4.1 Mini | 42.5 |
| 14 | DeepSeek V3 (0324) | 42.5 |
| 15 | GPT-4o | 41 |
Interactive version: theaggregate.ai/benchmark?slug=researchcodebench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.