GAIA — Level 3 difficulty — GAIA public leaderboard, Level 3 (hardest) split.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | DeepSeek-V3.2 | DeepSeek | 61.2 | 1 | 2026-05-10 |
| 2 | GPT-5.2 | OpenAI | 57.1 | 1 | 2026-05-10 |
| 3 | Gemini 3 Flash Preview | 49 | 1 | 2026-05-11 | |
| 4 | GPT-5 | OpenAI | 46.9 | 1 | 2026-05-10 |
| 5 | GPT-4o | OpenAI | 14.3 | 1 | 2026-05-10 |
| 6 | GPT-5 nano | OpenAI | 8.2 | 1 | 2026-05-10 |
| 7 | Gemma 4 26B A4B IT | 6.1 | 1 | 2026-05-16 | |
| 8 | Qwen2.5 7B | Alibaba | 2 | 1 | 2026-05-10 |
| 9 | GPT-4 Turbo | OpenAI | 2 | 1 | 2026-05-11 |
| 10 | GPT4 | OpenAI | 0 | 1 | 2026-05-10 |
| 11 | GPT-4o mini | OpenAI | 0 | 1 | 2026-05-10 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.