SEAL - SWE Atlas - Test Writing: leaderboard
Scale SEAL evaluation of LLM ability to write test suites for software engineering tasks.
Metric: Score. Source: scale.com. Saturation forecast: Around May 2028. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5.1 (xHigh) | 67.04 | #9 (Claude Fable 5.1) |
| 2 | Claude Opus 5 (xHigh) | 62.22 | #12 (Claude Opus 5) |
| 3 | Claude Fable 5 (xHigh) | 55.6 | #16 (Claude Fable 5) |
| 4 | Gemini 3.8 Flash | 53.7 | #21 |
| 5 | GPT-6 Astra (xHigh) | 50.74 | #6 (GPT-6 Astra) |
| 6 | Claude Opus 4.8 (xHigh) | 49.63 | #37 (Claude Opus 4.8) |
| 7 | GPT-5.6 Sol (xHigh) | 45.9 | #19 (GPT-5.6 Sol) |
| 8 | GPT-5.5 (xHigh) | 42.59 | #31 (GPT-5.5) |
| 9 | GLM-5.2 | 41.48 | #73 |
| 10 | Muse Spark 1.1 (xHigh) | 41.48 | #52 (Muse Spark 1.1) |
| 11 | GPT-5.4 (xHigh) | 40 | #79 (GPT-5.4) |
| 12 | GPT-5.3 Codex (xHigh) | 38.98 | #69 (GPT-5.3 Codex) |
| 13 | Claude Opus 4.7 | 38.52 | #49 |
| 14 | Claude Opus 4.6 | 36.08 | #63 |
| 15 | Claude Sonnet 4.6 | 31.76 | #90 |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=seal-swe-atlas-test-writing · How It Works · Data refreshed daily, snapshot 2026-10-20.