LLM Stats (DeepSWE 1.1): leaderboard
Software-engineering agent benchmark evaluating models run in the standardized mini-swe-agent harness on end-to-end software engineering tasks.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around January 2027. 49 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Muse Wonder | 78.8 | #14 |
| 2 | Gemini 4 Argon | 77.9 | #8 |
| 3 | DeepSeek V4.1 Flash (1017) | 77 | #20 |
| 4 | Muse Spark 1.3 | 75.4 | #27 |
| 5 | GPT-6.1 Sol | 75.22 | #7 |
| 6 | Ember-1 | 75.2 | #43 |
| 7 | DeepSeek V4.1 Flash | 74.2 | #33 |
| 8 | Claude Opus 5.5 | 74.2 | #4 |
| 9 | GPT-6 Astra | 74.1 | #6 |
| 10 | Gemini 3.8 Flash | 73.7 | #23 |
| 11 | MiniMax-M3.1-Flash | 73.7 | #39 |
| 12 | GPT-5.6 Sol | 73 | #18 |
| 13 | MiMo-V2.6-Pro | 71.9 | #41 |
| 14 | Claude Sonnet 5.5 | 71 | #11 |
| 15 | Grok 4.7 | 71 | #26 |
No result here: #9 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-deepswe-1-1 · How It Works · Data refreshed daily, snapshot 2026-10-23.