Senior SWE-Bench: leaderboard
Senior SWE-Bench (Snorkel) — senior-engineer tasks from production PRs; tasteful solve rate combines verifier passes, expert rubrics and a code-quality taste judge. Rows preserve the primary-selected API model, reasoning effort and agent harness; reported denominators vary from 93 to 95 in the reviewed roster.
Metric: Tasteful Solve Rate (pass@1, %). Source: senior-swe-bench.snorkel.ai. Saturation forecast: Around August 2028. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (High) | 39.33 | #3 (Claude Opus 5.5) |
| 2 | GPT-6 Astra (High) | 36.84 | #6 (GPT-6 Astra) |
| 3 | GPT-5.6 Sol (xHigh) | 34.74 | #19 (GPT-5.6 Sol) |
| 4 | Claude Fable 5 (High) | 34.74 | #16 (Claude Fable 5) |
| 5 | Claude Fable 5.1 (Medium) | 34.74 | #9 (Claude Fable 5.1) |
| 6 | Claude Opus 4.8 (Max) | 30.53 | #37 (Claude Opus 4.8) |
| 7 | GPT-5.5 (xHigh) | 29.47 | #31 (GPT-5.5) |
| 8 | Claude Sonnet 5 (Max) | 27.96 | #57 (Claude Sonnet 5) |
| 9 | Grok 4.7 (xHigh) | 27.37 | #25 (Grok 4.7) |
| 10 | GPT-5.6 Terra (High) | 27.37 | #41 (GPT-5.6 Terra) |
| 11 | Grok 4.6 (High) | 26.32 | #26 (Grok 4.6) |
| 12 | Grok 4.5 (Medium) | 23.16 | #48 (Grok 4.5) |
| 13 | DeepSeek V4.1 Flash (High) | 22.67 | #36 (DeepSeek V4.1 Flash) |
| 14 | Kimi K3 (Max) | 20.21 | #30 (Kimi K3) |
| 15 | Claude Opus 4.7 (Max) | 18.09 | #49 (Claude Opus 4.7) |
No result here: #7 GPT-6.1 Sol, #8 Gemini 4 Argon, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=senior-swe-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.