GAIA — text-only modality restriction — GAIA restricted to text-only inputs — distinct from no_file/level splits.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | GPT-5 | OpenAI | 76.4 | 1 | 2026-06-06 |
| 2 | MiniMax M2 | MiniMax | 75.7 | 1 | 2026-06-06 |
| 3 | GLM-4.6 | Z.ai | 71.9 | 1 | 2026-06-06 |
| 4 | Claude Sonnet 4.5 | Anthropic | 71.2 | 1 | 2026-06-06 |
| 5 | Claude Sonnet 4 | Anthropic | 68.3 | 1 | 2026-06-06 |
| 6 | DeepSeek-V3.2 | DeepSeek | 63.5 | 1 | 2026-06-06 |
| 7 | Kimi K2 (Reasoning) | Moonshot | 60.2 | 1 | 2026-06-06 |
| 8 | Gemini 2.5 Pro | 60.2 | 1 | 2026-06-06 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.