Benchmarks · agentic
Scale SEAL: SWE-Bench Pro V2 HARD
Scale SEAL: SWE-Bench Pro V2 HARD results published by Scale AI. Maintained by Scale AI. Score: Scale SEAL benchmark score.
Category
11 models · 11 results · data as of 2026-10-29 · scale.com ↗
Leaderboard
Opus 5 xhigh
All results
11 rows| # | Model | Effort | Score | Harness |
|---|---|---|---|---|
| 1 | Opus 5 | xhigh | 98.0% | |
| 2 | Fable 5.1 | high | 92.2% | |
| 3 | GPT-6 Astra Codex | high | 90.2% | Codex |
| 4 | Sonnet 5 | xhigh | 88.2% | |
| 5 | Kimi K3 | max | 88.2% | |
| 6 | GPT-5.6 Terra Codex | xhigh | 86.3% | |
| 7 | GLM 5.3 | max | 84.3% | |
| 8 | GPT-5.6 Sol Codex | xhigh | 82.4% | |
| 9 | Gemini 3.8 Flash | high | 58.8% | |
| 10 | Inkling | xhigh | 56.9% | |
| 11 | Haiku 4.5 | xhigh | 25.5% |