BLUEX v2 (Official): leaderboard
Metric: Mean rubric score (0-10; 880 subquestions; evaluation errors score zero). Source: github.com. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 9.1 |
| 2 | GPT-5.4 | 8.92 |
| 3 | GPT-5 Mini | 8.9 |
| 4 | Qwen 3.5 122B A10B | 8.83 |
| 5 | Gemini 3.1 Flash Lite (Preview) | 8.83 |
| 6 | Qwen 3.5 35B A3B | 8.77 |
| 7 | Qwen 3.5 27B | 8.74 |
| 8 | Qwen 3.5 9B | 8.63 |
| 9 | Gemma 4 26B A4B (IT) | 8.6 |
| 10 | Qwen 3 14B | 8.57 |
| 11 | Qwen 3 30B A3B 2507 Instruct | 8.54 |
| 12 | Mistral Small 4 | 8.43 |
| 13 | Qwen 3 8B | 8.41 |
| 14 | Llama 4 Scout | 8.1 |
| 15 | Phi-4 | 7.09 |
Interactive version: theaggregate.ai/benchmark?slug=bluex-v2-official · How It Works · Data refreshed daily, snapshot 2026-10-02.