LLaVA-Bench: leaderboard
LLaVA-Bench: Evaluates multimodal understanding across image, text, chart, diagram, or cross-modal reasoning tasks.
Metric: Score. Source: github.com. Source: huggingface.co. Saturation forecast: Estimated already saturated. 234 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 105.6 | #250 |
| 2 | Gemma 3 27B | 103.1 | #607 |
| 3 | InternVL3-78B | 100.4 | #352 |
| 4 | Grok 2 (1212) | 100 | #496 |
| 5 | GPT-4.1 Mini | 99.6 | #355 |
| 6 | Gemini 2.0 Flash | 97.7 | #342 |
| 7 | Gemini 1.5 Pro | 95.3 | #445 |
| 8 | InternVL3-38B | 94.7 | #405 |
| 9 | Claude 3.7 Sonnet | 94.4 | #249 |
| 10 | Gemma 3 12B | 93 | #671 |
| 11 | Gemma 3 4B | 92.5 | #1089 |
| 12 | Qwen 2.5 VL 7B | 91 | #643 |
| 13 | Claude 3.5 Sonnet (20241022) | 89.4 | #291 |
| 14 | GPT-4.1 Nano | 88.8 | #731 |
| 15 | Llama 3.2 90B Vision Instruct | 86.5 | #622 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llava-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.