Apple silicon (unified memory)
What AI models can the Apple M4 run locally?
At 4-bit quantization with an 8K-token conversation, 60 of the 146 open-weight models in our directory fit entirely in the Apple M4's 32 GB of unified memory (about 24 GB usable, the default GPU share, often adjustable), up to Qwen3.5 35B-A3B. Among models of 7B parameters or more, Laguna XS 2.1 (free) is the fastest, at an estimated 46 to 59 tokens per second.
| Model | Parameters | Memory needed | Estimated speed |
|---|---|---|---|
| Runs entirely in fast memory (60) | |||
| Qwen3.5 35B-A3B Alibaba (Qwen) | 36.0B 4.7B active | 23 GB 22 GB + 0.2 GB cache | 24–31 tokens/s Comfortable to read, little memory to spare |
| Qwen3.6 35B-A3B Alibaba (Qwen) | 36.0B 4.7B active | 23 GB 22 GB + 0.2 GB cache | 24–31 tokens/s Comfortable to read, little memory to spare |
| Nex-N2.5-Mini NEX AGI | 35.1B 3.9B active | 22 GB 21 GB + 0.2 GB cache | 29–38 tokens/s Fast, little memory to spare |
| Laguna XS 2.1 (free) Poolside | 33.4B 2.2B active | 21 GB 20 GB + 0.4 GB cache | 46–59 tokens/s Fast |
| Qwen3 VL 32B Instruct Alibaba (Qwen) | 33.4B | 23 GB 20 GB + 2.1 GB cache | 3.4–4.3 tokens/s Too slow for chat, little memory to spare |
| Qwen2.5 Coder 32B Instruct Alibaba (Qwen) | 32.8B | 23 GB 20 GB + 2.1 GB cache | 3.4–4.4 tokens/s Too slow for chat, little memory to spare |
| Qwen3 32B Alibaba (Qwen) | 32.8B | 23 GB 20 GB + 2.1 GB cache | 3.4–4.4 tokens/s Too slow for chat, little memory to spare |
| Nemotron 3 Nano 30B A3B NVIDIA | 31.6B | 20 GB 19 GB + 0.1 GB cache | Speed not estimated for this architecture |
| Nemotron 3.5 Lightning 30B A3B NVIDIA | 31.6B | 20 GB 19 GB + 0.4 GB cache | Speed not estimated for this architecture |
| Gemma 4 31B IT Google DeepMind | 31.3B | 22 GB 19 GB + 2.2 GB cache | 3.6–4.6 tokens/s Usable but slow |
| GLM-4.7-Flash Zhipu AI (Z.ai) | 31.2B 5.2B active | 20 GB 19 GB + 0.4 GB cache | 21–28 tokens/s Comfortable to read |
| Qwen3 VL 30B A3B Instruct Alibaba (Qwen) | 31.1B 3.9B active | 20 GB 19 GB + 0.8 GB cache | 26–33 tokens/s Comfortable to read |
| Qwen3 VL 30B A3B Thinking Alibaba (Qwen) | 31.1B 3.9B active | 20 GB 19 GB + 0.8 GB cache | 26–33 tokens/s Comfortable to read |
| Qwen3-Coder 30B-A3B Instruct Alibaba (Qwen) | 30.5B 3.4B active | 20 GB 19 GB + 0.8 GB cache | 29–38 tokens/s Fast |
| Qwen3 30B A3B Instruct 2507 Alibaba (Qwen) | 30.5B 3.4B active | 20 GB 19 GB + 0.8 GB cache | 29–38 tokens/s Fast |
| Qwen3 30B A3B Thinking 2507 Alibaba (Qwen) | 30.5B 3.4B active | 20 GB 19 GB + 0.8 GB cache | 29–38 tokens/s Fast |
| Qwen3 30B A3B Alibaba (Qwen) | 30.5B 3.4B active | 20 GB 19 GB + 0.8 GB cache | 29–38 tokens/s Fast |
| Hy-MT2-30B-A3B Tencent (Hunyuan) | 30.1B 3.5B active | 20 GB 18 GB + 0.8 GB cache | 29–37 tokens/s Fast |
| Muse Glimmer 30B Meta | 29.8B | 19 GB 18 GB + 0.2 GB cache | 3.9–5.1 tokens/s Usable but slow |
| Qwen3.8 27B Alibaba (Qwen) | 27.8B | 18 GB 17 GB + 0.5 GB cache | 4.2–5.4 tokens/s Usable but slow |
| Qwen3.5 27B Alibaba (Qwen) | 27.8B | 18 GB 17 GB + 0.5 GB cache | 4.2–5.4 tokens/s Usable but slow |
| Qwen3.6 27B Alibaba (Qwen) | 27.8B | 18 GB 17 GB + 0.5 GB cache | 4.2–5.4 tokens/s Usable but slow |
| Gemma 3 27B IT Google DeepMind | 27.4B | 17 GB weights only | Speed not estimated (architecture not public) |
| Gemma 2 27B Google DeepMind | 27.2B | 17 GB weights only | Speed not estimated (architecture not public) |
| Gemma 4 26B A4B IT Google DeepMind | 25.8B | 17 GB 16 GB + 0.5 GB cache | Speed not estimated for this architecture |
| Voxtral Small 24B 2507 Mistral AI | 24.3B | 17 GB 15 GB + 1.3 GB cache | 4.6–6.0 tokens/s Usable but slow |
| Uncensored Cognitivecomputations | 24.0B | 17 GB 15 GB + 1.3 GB cache | 4.7–6.1 tokens/s Usable but slow |
| Mistral Small 3.1 24B Instruct Mistral AI | 24.0B | 17 GB 15 GB + 1.3 GB cache | 4.7–6.1 tokens/s Usable but slow |
| LFM2.5-24B-A2B Liquid | 24.0B | 16 GB 15 GB + 0.2 GB cache | Speed not estimated for this architecture |
| Mistral Small 3 Mistral AI | 23.6B | 16 GB 14 GB + 1.3 GB cache | 4.8–6.2 tokens/s Usable but slow |
| Cydonia 24B V4.1 Thedrummer | 23.6B | 16 GB 14 GB + 1.3 GB cache | 4.8–6.2 tokens/s Usable but slow |
| Cydonia 24B V4.1 Thedrummer | 23.6B | 16 GB 14 GB + 1.3 GB cache | 4.8–6.2 tokens/s Usable but slow |
| Safety GPT OSS 20B OpenAI | 21.5B 4.8B active | 14 GB 13 GB + 0.2 GB cache | 24–31 tokens/s Comfortable to read |
| GPT OSS 20B OpenAI | 20.9B 4.2B active | 14 GB 13 GB + 0.2 GB cache | 27–35 tokens/s Fast |
| Reka Flash 3 Rekaai | 20.9B | 15 GB 13 GB + 1.1 GB cache | 5.4–7.0 tokens/s Usable but slow |
| Qwen3 14B Alibaba (Qwen) | 14.8B | 11 GB 9.0 GB + 1.3 GB cache | 7.4–9.6 tokens/s Usable but slow |
| Phi 4 Microsoft | 14.7B | 11 GB 8.9 GB + 1.7 GB cache | 7.3–9.5 tokens/s Usable but slow |
| Ministral 3 14B Mistral AI | 13.9B | 11 GB 8.5 GB + 1.3 GB cache | 7.8–10 tokens/s Usable but slow |
| Gemma 3 12B IT Google DeepMind | 12.2B | 8.2 GB weights only | Speed not estimated (architecture not public) |
| Llama Guard 4 12B Meta | 12.0B | 8.1 GB weights only | Speed not estimated (architecture not public) |
| Qwen3.5 9B Alibaba (Qwen) | 9.7B | 6.9 GB 5.9 GB + 0.3 GB cache | 12–15 tokens/s Comfortable to read |
| Moondream 3.1 Moondream | 9.3B | 6.4 GB weights only | Speed not estimated (architecture not public) |
| Moondream 3 Preview Moondream | 9.3B | 6.4 GB weights only | Speed not estimated (architecture not public) |
| Ministral 3 8B Mistral AI | 8.9B | 7.3 GB 5.4 GB + 1.1 GB cache | 12–15 tokens/s Comfortable to read |
| Granite 4.2 8B IBM | 8.8B | 7.5 GB 5.3 GB + 1.3 GB cache | 12–15 tokens/s Comfortable to read |
| Qwen3 VL 8B Instruct Alibaba (Qwen) | 8.8B | 7.3 GB 5.3 GB + 1.2 GB cache | 12–16 tokens/s Comfortable to read |
| Qwen3 VL 8B Thinking Alibaba (Qwen) | 8.8B | 7.3 GB 5.3 GB + 1.2 GB cache | 12–16 tokens/s Comfortable to read |
| UI-TARS 7B Bytedance | 8.3B | 6.3 GB 5.0 GB + 0.5 GB cache | 14–18 tokens/s Comfortable to read |
| Qwen3 8B Alibaba (Qwen) | 8.2B | 7.0 GB 5.0 GB + 1.2 GB cache | 13–17 tokens/s Comfortable to read |
| Llama 3.1 8B Instruct fp8 Meta | 8.0B | 5.7 GB weights only | Speed not estimated (architecture not public) |
| Hy-MT2-7B Tencent (Hunyuan) | 8.0B | 6.7 GB 4.9 GB + 1.1 GB cache | 13–17 tokens/s Comfortable to read |
| Qwen2.5 7B Instruct Alibaba (Qwen) | 7.6B | 5.9 GB 4.6 GB + 0.5 GB cache | 15–19 tokens/s Comfortable to read |
| Gemma 3 4B IT Google DeepMind | 4.3B | 3.4 GB weights only | Speed not estimated (architecture not public) |
| Nemotron 3.5 Content Safety (free) NVIDIA | 4.3B | 3.7 GB 2.6 GB + 0.3 GB cache | 26–34 tokens/s Comfortable to read |
| Ministral 3 3B Mistral AI | 3.8B | 4.0 GB 2.3 GB + 0.9 GB cache | 26–33 tokens/s Comfortable to read |
| Llama 3.2 3B Instruct Meta | 3.2B | 2.7 GB weights only | Speed not estimated (architecture not public) |
| Granite 4.0 Micro IBM | 3.2B | 2.8 GB 1.9 GB + 0.1 GB cache | 36–47 tokens/s Fast |
| Hy-MT2-1.8B Tencent (Hunyuan) | 2.0B | 2.6 GB 1.2 GB + 0.5 GB cache | 47–61 tokens/s Fast |
| Hy-MT2-1.8B Tencent (Hunyuan) | 2.0B | 2.6 GB 1.2 GB + 0.5 GB cache | 47–61 tokens/s Fast |
| Llama 3.2 1B Instruct Meta | 1.2B | 1.5 GB weights only | Speed not estimated (architecture not public) |
86 larger models do not fit. Change the quantization, conversation length or memory size in the tool.
Compared with similar chips
- Apple M5: same memory, so the same models fit; about 27% faster on models that fit both (153 vs 120 GB/s).
- Apple M4 Max (14-core CPU): 36 GB, 61 models fit entirely (+1); about 192% faster on models that fit both (410 vs 120 GB/s).
Frequently asked questions
What is the largest AI model the Apple M4 can run?
At 4-bit quantization with an 8K-token conversation, the largest model in our directory that fits entirely in its 32 GB of unified memory is Qwen3.5 35B-A3B (23 GB).
Can the Apple M4 run GPT OSS 20B?
Yes. At 4-bit, GPT OSS 20B needs about 14 GB and fits entirely in the Apple M4's memory at an estimated 27 to 35 tokens per second.
Can the Apple M4 run Qwen3 32B?
Yes. At 4-bit, Qwen3 32B needs about 23 GB and fits entirely in the Apple M4's memory at an estimated 3.4 to 4.4 tokens per second.
Can the Apple M4 run OpenAI GPT 120B?
No. OpenAI GPT 120B needs about 72 GB at 4-bit, more than the Apple M4 can hold.
These are estimates, not measurements: memory from each model's parameter count and published architecture, speed from memory bandwidth, calibrated on the AIPCs test bench. Drivers, software and prompt length move real results. How the estimate works.