MLX Benchmark V2 - MLX LM: leaderboard
MLX Benchmark V2 category score for MLX language-model workflows.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Not forecast. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 90.16 | #90 |
| 2 | Qwen 3.6 Max Preview | 80.56 | #85 |
| 3 | Gemini 3 Flash (Preview) | 78.69 | #82 |
| 4 | Gemma 4 26B A4B (IT) | 77.05 | #297 |
| 5 | DeepSeek V4 Flash | 73.77 | #137 |
| 6 | Grok 4.1 Fast | 72.13 | #217 |
| 7 | DeepSeek V4 Pro | 71.93 | #102 |
| 8 | GPT-5.4 Nano | 70.49 | #328 |
| 9 | Gemini 2.5 Flash Lite (Preview 09-2025) | 68.85 | #318 |
| 10 | GPT-5 Nano | 49.18 | #425 |
| 11 | Qwen 3.6 35B A3B | 49.18 | #226 |
| 12 | GLM-5.1 | 19.67 | #97 |
| 13 | Nemotron 3 Ultra | 18.03 | #235 |
| 14 | Kimi K2.5 | 8.2 | #144 |
| 15 | Kimi K2.6 | 4.92 | #104 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mlx-benchmark-v2-mlx-lm · How It Works · Data refreshed daily, snapshot 2026-10-20.