MLX Benchmark V2 - Debugging — leaderboard
Metric: Accuracy (%). Source: huggingface.co. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 95.24 |
| 2 | Gemini 3 Flash (Preview) | 80.95 |
| 3 | Grok 4.1 Fast | 71.43 |
| 4 | Gemma 4 26B A4B (IT) | 66.67 |
| 5 | GPT-5.4 Nano | 57.14 |
| 6 | Gemini 2.5 Flash Lite (Preview 09-2025) | 52.38 |
| 7 | DeepSeek V4 Pro | 33.33 |
| 8 | Qwen 3.6 35B A3B | 33.33 |
| 9 | Nemotron 3 Ultra | 19.05 |
| 10 | GPT-5 Nano | 14.29 |
| 11 | DeepSeek V4 Flash | 14.29 |
| 12 | Kimi K2.5 | 0 |
| 13 | GLM-5.1 | 0 |
| 14 | Kimi K2.6 | 0 |
Interactive version: theaggregate.ai/benchmark?slug=mlx-benchmark-v2-debugging · How the rankings work · Data refreshed daily, snapshot 2026-07-22.