OpenVLM LLaVA-Bench - Complex: leaderboard
OpenCompass OpenVLM LLaVA-Bench score for complex visual question answering prompts.
Metric: Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 234 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 27B | 112.4 | #607 |
| 2 | GPT-4.1 (2025-04-14) | 108.9 | #250 |
| 3 | Gemma 3 12B | 108.1 | #671 |
| 4 | GPT-4.1 Mini | 105.6 | #355 |
| 5 | Grok 2 (1212) | 105.3 | #496 |
| 6 | InternVL3-78B | 105.2 | #352 |
| 7 | Gemini 1.5 Pro | 101.9 | #445 |
| 8 | Claude 3.7 Sonnet | 101.3 | #249 |
| 9 | Gemini 2.0 Flash | 100.4 | #342 |
| 10 | Gemma 3 4B | 100 | #1089 |
| 11 | InternVL3-38B | 99.5 | #405 |
| 12 | GPT-4.1 Nano | 98.2 | #731 |
| 13 | InternVL3-8B | 97.7 | #620 |
| 14 | Qwen 2.5 VL 7B | 93.6 | #643 |
| 15 | InternVL3-14B | 90.1 | #505 |
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-complex · How It Works · Data refreshed daily, snapshot 2026-10-20.