What AI models run on the Mac M2 24GB?
The Mac M2 24GB has 24 GB of unified memory (100 GB/s). About 16.1 GB is usable for a model. 206 of 298 open models we track fit at 8K context.
Apple Silicon uses unified memory; see the Mac guide for default macOS limits across machines. Other 24 GB hardware: RTX 3090, RTX 3090 Ti, RTX 4090, RX 7900 XTX, Mac M3 24GB, Mac M4 Pro 24GB.
Using a specific tool? Ollama models that run on the Mac M2 24GB · LM Studio
Models that fit, largest first
Just too big at 4-bit
These need up to 35% more memory than the Mac M2 24GB has at Q4_K_M and 8K context. A smaller quant or shorter context may work.
| Model | Q4_K_M needs | Short by |
|---|---|---|
| GLiDE 27B | 16.2 GB | 0.1 GB |
| MiniMax H3 Image PE | 16.8 GB | 0.7 GB |
| DiffusionGemma 26B A4B IT | 17.1 GB | 1.0 GB |
| Gemma 4 26B A4B IT | 17.3 GB | 1.2 GB |
| Gemma 3 27B IT | 17.3 GB | 1.3 GB |
| Arete 27B | 17.4 GB | 1.3 GB |
| Qwen3.8 27B | 17.6 GB | 1.5 GB |
| Muse Glimmer 30B | 17.9 GB | 1.8 GB |
| Qwen3.5 27B | 18.1 GB | 2.0 GB |
| Qwen3.6 27B | 18.1 GB | 2.0 GB |
| GLM 4.7 Flash | 18.3 GB | 2.2 GB |
| Hy Memory 30B A3B | 18.5 GB | 2.4 GB |
| Qwen3 Coder 30B A3B Instruct | 18.8 GB | 2.7 GB |
| North Mini Code 1.0 | 18.8 GB | 2.8 GB |
| Granite 4.0 H Small | 18.9 GB | 2.8 GB |
Questions
What is the biggest model that runs on the Mac M2 24GB?
Of the models we track, the largest that fits at 8K context is Flint (40B) at Q2_K, needing about 15.4 GB.
How much of the Mac M2 24GB's memory can a model use?
About 16 GB of 24 GB. macOS lets the GPU use about 67% of unified memory by default; this can be raised with the iogpu.wired_limit_mb setting.
Spec source: Apple specifications. Model data: Hugging Face Hub API. Speed is a bandwidth ceiling (bandwidth ÷ bytes read per token), not a benchmark. Methodology · Report an error.