OpenVLM Leaderboard: leaderboard
OpenVLM Leaderboard evaluates vision-language models on multimodal understanding benchmarks such as MMBench, MMStar, MMMU, MathVista, OCRBench, AI2D, HallusionBench, and MMVet. The pinned metric is Avg Score, measuring the aggregated benchmark score in the mirrored OpenCompass/VLMEvalKit leaderboard tables. Source: BenchmarkList’s attributed result table.
Metric: Avg Score. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 80.1 | #159 |
| 2 | GPT-5 | 79.9 | #105 |
| 3 | InternVL3-78B | 79.1 | #361 |
| 4 | GPT-5 Mini (2025-08-07) | 78 | #175 |
| 5 | InternVL3-38B | 77.8 | #413 |
| 6 | GPT-4.1 (2025-04-14) | 75.9 | #257 |
| 7 | GPT-4o ChatGPT | 75.4 | |
| 8 | GPT-4.5 | 75.3 | #278 |
| 9 | InternVL3-14B | 75.2 | #506 |
| 10 | Qwen 2 VL 72B | 74.8 | #450 |
| 11 | InternVL3-8B | 73.6 | #624 |
| 12 | Gemini 2.0 Flash | 72.6 | #351 |
| 13 | GPT-5 Nano | 72.2 | #439 |
| 14 | Gemini 1.5 Pro (002) | 72.2 | #368 |
| 15 | Qwen 2.5 VL 7B | 70.9 | #641 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-leaderboard · How It Works · Data refreshed daily, snapshot 2026-11-05.