Eval Anything: leaderboard
PKU-Alignment Eval-Anything benchmark assessing all-modality models on an aggregate of vision, audio, and text understanding and instruction-following tasks.
Metric: Overall score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 1.5 Pro | 4.58 | #445 |
| 2 | GPT-4o | 3.92 | #343 |
| 3 | Qwen2-Audio-7B-Instruct | 2.73 | #880 |
| 4 | Qwen 2 VL 7B Instruct | 2.62 | #826 |
| 5 | Llama 3.1 8B Instruct | 2.14 | #1035 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=eval-anything · How It Works · Data refreshed daily, snapshot 2026-10-20.