# SWE-rebench

Tests AI coding tools on fresh real-world bug-fix tasks collected after models were trained, so they cannot have seen the answers.

Category: coding. Type: leaderboard. Status: active.
Metric: Share of tasks resolved on the first attempt (%) (higher is better).
Source: SWE-rebench (Nebius) — https://swe-rebench.com/, results as of 2026-07-01, retrieved 2026-11-17, license: not stated.

| Rank | Model / Agent | Resolved Rate (%) | Pass@5 (%) | Cost per Problem |
|---|---|---|---|---|
| 1 | Anthropic Fable 5 [high] | 64.5 | 78.4 | $4.40 |
| 2 | Grok Grok 4.5 [high] | 63.8 | 77.5 | $1.47 |
| 3 | Anthropic Opus 5 [high] | 63.4 | 74.8 | $3.47 |
| 4 | Z.ai GLM-5.2 [high] | 62.9 | 81.1 | $1.40 |
| 5 | OpenAI GPT-5.6 Sol [medium] | 62.3 | 79.3 | $0.85 |
| 6 | Junie Junie | 61.8 | 73.9 | $0.81 |
| 7 | Anthropic Claude Code | 60.4 | 75.7 | $3.39 |
| 8 | OpenAI Codex | 58.0 | 73.0 | $1.59 |
| 9 | Anthropic Sonnet 5 [high] | 56.8 | 74.8 | $1.43 |
| 10 | Cursor Cursor | 51.7 | 65.8 | $0.41 |
| 11 | Minimax MiniMax M3 | 47.2 | 69.4 | $0.95 |
| 12 | XiaomiMiMo MiMo V2.5 Pro | 46.5 | 65.8 | $0.10 |
| 13 | OpenAI GPT-5.6 Luna [medium] | 43.6 | 59.5 | $0.11 |
| 14 | DeepSeek DeepSeek-V4 Pro [high] | 40.2 | 64.0 | $0.15 |
| 15 | Qwen Qwen3.6-27B | 31.2 | 57.7 | $0.62 |
| 16 | Qwen Qwen3.6-35B-A3B | 24.7 | 43.2 | $0.27 |
| 17 | Qwen Qwen3.5-35B-A3B | 17.1 | 36.9 | $0.99 |

Notes: SWE-rebench v2, default time window 15/05/2026-01/07/2026 (111 problems). Entries without data in this window omitted. as_of is window end.
