MLX Benchmark V2 - MLX VLM — leaderboard
Metric: Accuracy (%). Source: huggingface.co. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 84.21 |
| 2 | Gemma 4 26B A4B (IT) | 78.95 |
| 3 | Claude Sonnet 4.6 | 73.68 |
| 4 | Grok 4.1 Fast | 73.68 |
| 5 | DeepSeek V4 Pro | 68.75 |
| 6 | GPT-5.4 Nano | 68.42 |
| 7 | DeepSeek V4 Flash | 63.16 |
| 8 | Qwen 3.6 Max Preview | 60 |
| 9 | Qwen 3.6 35B A3B | 57.89 |
| 10 | GPT-5 Nano | 47.37 |
| 11 | Gemini 2.5 Flash Lite (Preview 09-2025) | 47.37 |
| 12 | GLM-5.1 | 15.79 |
| 13 | Kimi K2.5 | 10.53 |
| 14 | Nemotron 3 Ultra | 10.53 |
| 15 | Kimi K2.6 | 5.26 |
Interactive version: theaggregate.ai/benchmark?slug=mlx-benchmark-v2-mlx-vlm · How the rankings work · Data refreshed daily, snapshot 2026-07-22.