Vals Index
Vals composite index · 10 models · source Vals AI · updated 2026-10
What this benchmark measures
A weighted score across vertical professional tasks — legal, finance, tax, biology, Excel modelling — all run by Vals itself, with no vendor-supplied numbers.
Why it gets its own table
Correlates with the AA index at about 0.9 but leans towards professional services. Models where the two disagree are worth a closer look — it usually means the training data has an industry bias.
Full ranking
| # | Model | Score | Vs. best | Input $ | Output $ | Type |
|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5Anthropic | 73.1 | Best overall | $2.00 | $10.00 | |
| 2 | Claude Opus 5.5Anthropic | 72.4 | 99% | $4.00 | $20.00 | |
| 3 | Gemini 4 ArgonGoogle | 70.36 | 96% | $2.00 | $10.00 | |
| 4 | GPT-6.1 TerraOpenAI | 60.47 | 83% | $1.00 | $5.00 | |
| 5 | MiMo-V2.6 Pro Open weightsXiaomi | 57.5 | 79% | $0.35 | $1.40 | Open weights |
| 6 | DeepSeek V4.1 Flash Open weightsDeepSeek | 50.9 | 70% | $0.30 | $1.20 | Open weights |
| 7 | Qwen3.8 27B Open weightsAlibaba Qwen | 40.5 | 55% | $0.10 | $0.40 | Open weights |
| 8 | MiniMax M3 Open weightsMiniMax | 29.1 | 40% | $0.25 | $1.00 | Open weights |
| 9 | Gemma 4 31B Open weightsGoogle | 16.8 | 23% | $0.10 | $0.40 | Open weights |
| 10 | gpt-oss-120b Open weightsOpenAI | 14.5 | 20% | $0.05 | $0.20 | Open weights |