Vals AI Code Migration v1.1: leaderboard
Code Migration v1.1: Pi-agent offline migration with revised hidden tests, distinct from v1 mini-SWE-agent runs. 30 CLI repositories each migrated into four target languages and 10 COBOL-to-Java repositories; source repositories equally weighted (3:1 CLI/COBOL). Anti-cheat validation; headline hidden-test pass rate excludes separate style ratings.
Metric: Hidden-test Pass Rate (%). Source: www.vals.ai. Saturation forecast: Around January 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 5.5 (Max) | 86.88 | #11 (Claude Sonnet 5.5) |
| 2 | Claude Opus 5.5 (Max) | 81.01 | #4 (Claude Opus 5.5) |
| 3 | GPT-6 Astra (Max) | 80.58 | #6 (GPT-6 Astra) |
| 4 | GPT-6.1 Sol (Max) | 77.32 | #8 (GPT-6.1 Sol) |
| 5 | Muse Wonder (Max) | 72.36 | #14 (Muse Wonder) |
| 6 | Gemini 4 Argon (High) | 70.8 | #7 (Gemini 4 Argon) |
| 7 | DeepSeek V4.1 Flash (1017) (Max) | 67.52 | #20 (DeepSeek V4.1 Flash (1017)) |
| 8 | Muse Wonder Full (Max) | 62.79 | #10 (Muse Wonder Full) |
| 9 | DeepSeek V4.1 Flash (Max) | 58.94 | #36 (DeepSeek V4.1 Flash) |
| 10 | GLM-5.3 (Max) | 55.07 | #39 (GLM-5.3) |
| 11 | Grok 4.7 (xHigh) | 53.22 | #27 (Grok 4.7) |
| 12 | Claude Haiku 5.5 (Max) | 52.36 | #40 (Claude Haiku 5.5) |
| 13 | DeepSeek V4 Pro (0813) (Max) | 51.99 | #53 (DeepSeek V4 Pro (0813)) |
| 14 | GLM-5.3 Flash (Max) | 51.65 | #67 (GLM-5.3 Flash) |
| 15 | MiMo-V2.6-Pro | 48.79 | #42 |
No result here: #9 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-code-migration-v1-1 · How It Works · Data refreshed daily, snapshot 2026-10-21.