Vals AI SWE-bench Multimodal: leaderboard
Resolution rate on 480 visual JavaScript repository issues using mini-swe-agent. Screenshots are attached before the run, internet access is disabled, and the repository test harness grades the final patch. Network-enabled diagnostic results are separate from this ranking.
Metric: Resolved (%). Source: www.vals.ai. Saturation forecast: Around January 2028. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 70.21 | #3 |
| 2 | Claude Fable 5.1 | 61.25 | #9 |
| 3 | Claude Sonnet 5.5 | 58.33 | #10 |
| 4 | Gemini 4 Argon | 56.04 | #7 |
| 5 | Muse Wonder Full | 47.29 | #11 |
| 6 | DeepSeek V4.1 Flash | 46.46 | #36 |
| 7 | Muse Wonder | 45.42 | #16 |
| 8 | Claude Haiku 5.5 | 43.96 | #35 |
| 9 | GPT-6.1 Sol | 43.54 | #8 |
| 10 | GPT-6 Astra | 43.33 | #5 |
| 11 | Muse Spark 1.3 (Max) | 43.12 | #25 (Muse Spark 1.3) |
| 12 | Grok 4.7 | 42.92 | #27 |
| 13 | MiMo-V2.6-Pro | 41.25 | #37 |
| 14 | Gemini 3.8 Flash | 40.83 | #21 |
| 15 | Step 5 Preview | 40.83 | #41 |
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-swe-bench-multimodal · How It Works · Data refreshed daily, snapshot 2026-10-16.