GAIA — leaderboard
General AI Assistants benchmark - 466 questions requiring multi-step reasoning, web browsing, and tool use. Questions are easy for humans (~92%) but hard for AI. Three difficulty levels.
Metric: Accuracy (%). Source: huggingface.co. Status: saturation imminent. 3460 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 77.41 |
| 2 | DeepSeek V3.2 | 76.74 |
| 3 | GPT-5 | 66.45 |
| 4 | Gemini 3.1 Pro (Preview) | 55.48 |
| 5 | Gemini 3 Flash | 41.53 |
| 6 | GPT-4o (2024-08-06) | 28.9 |
| 7 | DeepSeek V4 Flash | 23.59 |
| 8 | GPT-5 Nano | 23.26 |
| 9 | GPT-5.1 | 19.6 |
| 10 | GPT-4 Turbo | 11.3 |
| 11 | Gemma 4 26B A4B (IT) | 8.31 |
| 12 | GPT-4o Mini (2024-07-18) | 4.65 |
| 13 | GPT-4 | 3.99 |
| 14 | GPT-3.5 | 2.66 |
| 15 | Qwen 2.5 7B | 0.66 |
Interactive version: theaggregate.ai/benchmark?slug=gaia · How the rankings work · Data refreshed daily, snapshot 2026-07-22.