IDE-Bench: leaderboard
IDE coding-agent benchmark with 80 interactive software-engineering tasks that test whether models can make a correct fix on the first attempt. AfterQuery took the repository down and published no replacement, so the link is a January 2026 Wayback snapshot of the board.
Metric: Pass@1 Accuracy (imported). Source: web.archive.org. Source: benchmarklist.com. Saturation forecast: Around March 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 87.5 | #142 |
| 2 | GPT-5.2 | 85 | #110 |
| 3 | Claude Opus 4.5 | 83.75 | #83 |
| 4 | Claude Haiku 4.5 | 78.75 | #280 |
| 5 | GPT-5.1 Codex Max | 73.75 | #100 |
| 6 | Qwen 3 Max | 65 | #208 |
| 7 | Gemini 3 Pro | 55 | #80 |
| 8 | Grok 4.1 Fast | 35 | #217 |
| 9 | DeepSeek V3 | 31.25 | #321 |
| 10 | DeepSeek R1 | 20 | #251 |
| 11 | Grok Code Fast 1 | 11.25 | #337 |
| 12 | Llama 4 Maverick | 2.5 | #464 |
| 13 | Llama 4 Scout | 2.5 | #656 |
| 14 | Command-R+ | 0 | #891 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ide-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.