| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.8 | Anthropic | 71.7 | 2 | 2026-08-13 |
| 2 | DeepSeek-V4-Flash-Vision-Exp | DeepSeek | 63.6 | 1 | 2026-08-23 |
| 3 | Kimi K3 | Moonshot | 63 | 1 | 2026-08-13 |
| 4 | DeepSeek-V4-Flash | DeepSeek | 59.6 | 6 | 2026-08-23 |
| 5 | GLM-5.2 Full Open Source | Z.ai | 54.5 | 2 | 2026-08-13 |
| 6 | DeepSeek-V4-Pro | DeepSeek | 31.1 | 2 | 2026-08-13 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.