OpenVLM LLaVA-Bench - Conv: leaderboard
OpenCompass OpenVLM LLaVA-Bench score for conversational visual instruction-following prompts.
Metric: Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 234 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 102 | #250 |
| 2 | InternVL3-78B | 101.4 | #352 |
| 3 | Llama 3.2 90B Vision Instruct | 97.9 | #622 |
| 4 | Gemini 2.0 Flash | 96.7 | #342 |
| 5 | Grok 2 (1212) | 96.6 | #496 |
| 6 | InternVL3-38B | 95.2 | #405 |
| 7 | Qwen 2.5 VL 7B | 94.7 | #643 |
| 8 | Claude 3.5 Sonnet (20241022) | 93.8 | #291 |
| 9 | GPT-4.1 Mini | 93.5 | #355 |
| 10 | Llama 3.2 11B Instruct | 91.5 | #1126 |
| 11 | Qwen 2 VL 72B | 91.2 | #460 |
| 12 | Gemini 1.0 Pro | 90.8 | #933 |
| 13 | Gemma 3 27B | 89.6 | #607 |
| 14 | Gemini 1.5 Pro | 88.6 | #445 |
| 15 | Qwen 2 VL 7B | 87.5 | #750 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-llava-bench-conv · How It Works · Data refreshed daily, snapshot 2026-10-20.