Vals AI SWE-bench Verified: leaderboard
Vals AI SWE-bench Verified leaderboard, measuring the percentage of real GitHub issues resolved by coding agents.
Metric: Resolved (%). Source: www.vals.ai. Saturation forecast: Around January 2027. 88 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 97 | #12 |
| 2 | DeepSeek V4 Pro (0813) (Max) | 96.4 | #53 (DeepSeek V4 Pro (0813)) |
| 3 | GPT-5.6 Sol (Max) | 96.2 | #19 (GPT-5.6 Sol) |
| 4 | Grok 4.6 (High) | 95.6 | #26 (Grok 4.6) |
| 5 | GLM-5.3 (Max) | 95.4 | #39 (GLM-5.3) |
| 6 | GPT-5.6 Terra (Max) | 95.4 | #41 (GPT-5.6 Terra) |
| 7 | Claude Fable 5 (Max) | 95 | #16 (Claude Fable 5) |
| 8 | Kimi K3 | 93.4 | #30 |
| 9 | GPT-5.6 Luna (Max) | 93 | #68 (GPT-5.6 Luna) |
| 10 | GLM-5.3 Flash (Max) | 92 | #65 (GLM-5.3 Flash) |
| 11 | DeepSeek V4 Flash (0731) (High) | 88.8 | #78 (DeepSeek V4 Flash (0731)) |
| 12 | Claude Opus 4.8 (Max) | 88.6 | #37 (Claude Opus 4.8) |
| 13 | Grok 4.5 (High) | 86.6 | #48 (Grok 4.5) |
| 14 | Muse Spark 1.2 (xHigh) | 86.6 | #62 (Muse Spark 1.2) |
| 15 | Qwen 3.8 27B (xHigh) | 86 | #105 (Qwen 3.8 27B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-swe-bench-verified · How It Works · Data refreshed daily, snapshot 2026-10-20.