MLX Benchmark V2 - Medium: leaderboard
MLX Benchmark V2 medium split measuring Apple MLX framework knowledge on intermediate questions.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Not forecast. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 85.8 | #82 |
| 2 | Claude Sonnet 4.6 | 85.08 | #90 |
| 3 | Gemma 4 26B A4B (IT) | 82.32 | #297 |
| 4 | Grok 4.1 Fast | 78.45 | #217 |
| 5 | DeepSeek V4 Pro | 77.09 | #102 |
| 6 | Qwen 3.6 Max Preview | 76.67 | #85 |
| 7 | GPT-5.4 Nano | 75.14 | #328 |
| 8 | Gemini 2.5 Flash Lite (Preview 09-2025) | 71.82 | #318 |
| 9 | DeepSeek V4 Flash | 71.27 | #137 |
| 10 | GPT-5 Nano | 49.17 | #425 |
| 11 | Qwen 3.6 35B A3B | 46.41 | #226 |
| 12 | GLM-5.1 | 18.78 | #97 |
| 13 | Nemotron 3 Ultra | 17.13 | #235 |
| 14 | Kimi K2.5 | 4.97 | #144 |
| 15 | Kimi K2.6 | 2.76 | #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-medium · How It Works · Data refreshed daily, snapshot 2026-10-20.