Macs
Which AI models run on the M3 Max (48 GB, 40-core GPU)?
The M3 Max (48 GB, 40-core GPU) has 48 GB of memory, about 36 GB of it usable for a model, and 400 GB/s of memory bandwidth. 88 of the 110 models in our catalog run on it at 4-bit or better.
Data as of 2 November 2026 · Vendor specs
48 GBmemory
~36 GB usable for a model
400 GB/sbandwidth, sets the speed
88models run at 4-bit or better
Models that run on it
| Model | Best build | Memory | Speed |
|---|---|---|---|
| Llama 3.3 Nemotron Super 49B49B | Q4_K_M | 31 GB | ~7.4 tok/s slow our est. |
| Qwen3.6-35B-A3B35B MoE | Q4_K_M | 24 GB | ~109 tok/s ModelFit est. |
| Gemma 4 31B30.7B | Q4_K_M | 21 GB | ~16 tok/s ModelFit est. |
| Qwen3-Coder-30B-A3B30.5B MoE | Q8_0 | 34 GB | ~52 tok/s ModelFit est. |
| Nemotron 3 Nano 30B-A3B30B MoE | Q8_0 | 34 GB | ~69 tok/s our est. |
| Qwen3.8-27B27.3B | Q8_0 | 29 GB | ~12 tok/s ModelFit est. |
| Gemma 4 26B-A4B25.2B MoE | Q8_0 | 30 GB | ~49 tok/s ModelFit est. |
| gpt-oss-20b21B MoE | MXFP4 | 15.1 GB | ~72 tok/s ModelFit est. |
| Nemotron Nano 9B v29B | FP16 | 19.4 GB | ~12 tok/s our est. |
| Qwen3.5 9B Instruct9B | Q8_0 | 12.0 GB | ~36 tok/s ModelFit est. |
| Qwen2.5 7B Instruct7B | Q4_K_M | 6.6 GB | ~64 tok/s ModelFit est. |
| Gemma 3 4B Instruct4B | Q8_0 | 5.0 GB | ~67 tok/s ModelFit est. |
| Mixtral 8x7B Instruct46.7B MoE | Q4_K_M | 32 GB | ~25 tok/s ModelFit est. |
| Flint 40B-A2.5B39.8B MoE | NVFP4 | 24 GB | ~118 tok/s ModelFit est. |
| Sage-A1.535.95B | Q4_K_M | 23 GB | ~109 tok/s ModelFit est. |
| Qwen3.5 35B-A3B Instruct35B MoE | Q4_K_M | 21 GB | ~111 tok/s ModelFit est. |
| Ornith 1.0 35B35B | Q4_K_M | 23 GB | ~111 tok/s ModelFit est. |
| Nanbeige4.3 33B33.1B | Q4_K_M | 22 GB | ~15 tok/s ModelFit est. |
| Poolside Laguna XS.233B | Q4_K_M | 25 GB | no estimate |
| Laguna XS 2.133B | Q4_K_M | 22 GB | ~102 tok/s ModelFit est. |
| Qwen3 32B32B | Q4_K_M | 22 GB | ~15 tok/s ModelFit est. |
| DeepSeek-R1 Distill Qwen 32B32B | Q4_K_M | 21 GB | ~16 tok/s ModelFit est. |
| Olmo 3.1 32B Instruct32B | Q4_K_M | 21 GB | ~16 tok/s ModelFit est. |
| Qwen3 30B30B | Q8_0 | 32 GB | ~52 tok/s ModelFit est. |
| NVIDIA Nemotron Cascade 2 30B-A3B30B MoE | Q6_K | 26 GB | ~92 tok/s our est. |
| Cohere North Mini Code30B | Q4_K_M | 20 GB | no estimate |
| Nemotron 3.5 Lightning 30B-A3B30B MoE | Q4_K_M | 27 GB | ~68 tok/s ModelFit est. |
| GLM-4.7-Flash 30B-A3B30B MoE | Q4_K_M | 20 GB | ~76 tok/s ModelFit est. |
| Granite Switch 4.1 30B30B | Q4_K_M | 23 GB | ~14 tok/s ModelFit est. |
| Muse Glimmer 30B29.8B | Q4_K_M | 18.3 GB | ~20 tok/s ModelFit est. |
| Granite 4.2 30B29.3B | Q4_K_M | 17.9 GB | ~17 tok/s ModelFit est. |
| AREX-2 27B27.4B | Q4_K_M | 18.6 GB | ~19 tok/s ModelFit est. |
| Gemma 2 27B Instruct27B | Q4_K_M | 23 GB | ~17 tok/s ModelFit est. |
| Gemma 3 27B Instruct27B | Q4_K_M | 23 GB | ~18 tok/s ModelFit est. |
| Qwen3.5 27B Instruct27B | Q8_0 | 29 GB | ~12 tok/s ModelFit est. |
| Qwen3.6 27B27B | Q8_0 | 32 GB | ~12 tok/s ModelFit est. |
| Jebadiah 27B27B | Q8_0 | 31 GB | ~12 tok/s ModelFit est. |
| Mistral Small 3.124B | Q8_0 | 25 GB | ~12 tok/s ModelFit est. |
| LFM2 24B-A2B Instruct24B MoE | Q4_K_M | 15.3 GB | ~119 tok/s ModelFit est. |
| Devstral Small 2 24B24B | Q4_K_M | 16.6 GB | ~20 tok/s ModelFit est. |
| Mistral Small 22B22B | Q4_K_M | 18.4 GB | ~23 tok/s ModelFit est. |
| Qwen2.5 14B Instruct14B | Q4_K_M | 12.3 GB | ~33 tok/s ModelFit est. |
| Phi-3 Medium 14B14B | Q4_K_M | 12.3 GB | ~20 tok/s our est. |
| Phi-4 14B14B | Q8_0 | 15.9 GB | ~20 tok/s ModelFit est. |
| Qwen2.5 Coder 14B14B | Q4_K_M | 12.3 GB | ~33 tok/s ModelFit est. |
| Qwen3 14B14B | Q8_0 | 17.3 GB | ~20 tok/s ModelFit est. |
| DeepSeek-R1 Distill Qwen 14B14B | Q8_0 | 16.0 GB | ~20 tok/s ModelFit est. |
| Ministral 3 14B13.5B | Q4_K_M | 10.3 GB | ~33 tok/s ModelFit est. |
| Mellum2.1 12B-A2.5B Thinking12.1B MoE | Q4_K_M | 9.3 GB | ~131 tok/s our est. |
| Mistral Nemo 12B12B | Q8_0 | 13.4 GB | ~24 tok/s ModelFit est. |
| Gemma 3 12B Instruct12B | Q4_K_M | 10.7 GB | ~39 tok/s ModelFit est. |
| Gemma 4 12B12B | Q8_0 | 14.1 GB | ~24 tok/s ModelFit est. |
| GLM-4-9B-Chat9.4B | Q8_0 | 11.2 GB | ~33 tok/s ModelFit est. |
| Gemma 2 9B Instruct9B | Q4_K_M | 8.2 GB | ~42 tok/s ModelFit est. |
| Ornith 1.0 9B9B | Q4_K_M | 6.7 GB | ~59 tok/s ModelFit est. |
| MiMo-V2.6-Distill-Qwen-9B9B | Q4_K_M | 6.9 GB | ~57 tok/s ModelFit est. |
| Jebadiah 9B9B | Q8_0 | 11.0 GB | ~35 tok/s ModelFit est. |
| Granite 4.2 8B8.8B | Q4_K_M | 6.1 GB | ~52 tok/s ModelFit est. |
| Granite Switch 4.1 8B8.8B | Q4_K_M | 7.6 GB | ~43 tok/s ModelFit est. |
| MiniCPM-V 4.5 8B8.7B | Q4_K_M | 6.8 GB | ~57 tok/s ModelFit est. |
| Ministral 3 8B8.4B | Q4_K_M | 7.1 GB | ~48 tok/s ModelFit est. |
| LFM2.5 8B-A1B8.3B MoE | Q4_K_M | 6.6 GB | ~152 tok/s ModelFit est. |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.2 GB | ~36 tok/s ModelFit est. |
| Qwen3 8B8B | Q8_0 | 9.3 GB | ~35 tok/s ModelFit est. |
| Granite 4.1 8B Instruct8B | Q4_K_M | 6.6 GB | ~50 tok/s ModelFit est. |
| Mistral 7B Instruct7B | Q8_0 | 8.4 GB | ~38 tok/s ModelFit est. |
| Qwen2.5 Coder 7B7B | Q4_K_M | 6.6 GB | ~64 tok/s ModelFit est. |
| DeepSeek-R1 Distill Qwen 7B7B | Q8_0 | 8.7 GB | ~40 tok/s ModelFit est. |
| Gemma 4 E4B4.5B | Q8_0 | 8.7 GB | ~59 tok/s ModelFit est. |
| Qwen3.5 4B Instruct4B | Q8_0 | 5.4 GB | ~64 tok/s ModelFit est. |
| NVIDIA Nemotron 3.5 Bolt E4B4B | Q4_K_M | 9.1 GB | ~74 tok/s ModelFit est. |
| Phi-3 Mini 3.8B3.8B | Q4_K_M | 4.3 GB | ~69 tok/s our est. |
| Phi-4 Mini 3.8B3.8B | Q8_0 | 4.9 GB | ~64 tok/s ModelFit est. |
| Granite 4.2 3B3.7B | Q4_K_M | 3.2 GB | ~112 tok/s ModelFit est. |
| Ministral 3 3B3.4B | Q4_K_M | 4.1 GB | ~84 tok/s ModelFit est. |
| Qwen2.5 3B Instruct3B | Q4_K_M | 3.6 GB | ~127 tok/s ModelFit est. |
| Llama 3.2 3B Instruct3B | Q4_K_M | 3.6 GB | ~109 tok/s ModelFit est. |
| Granite 4.1 3B Instruct3B | Q4_K_M | 3.0 GB | ~111 tok/s ModelFit est. |
| Gemma 4 E2B2.3B | Q8_0 | 5.7 GB | ~127 tok/s ModelFit est. |
| Gemma 2 2B Instruct2B | Q4_K_M | 2.8 GB | ~123 tok/s ModelFit est. |
| Qwen3.5 2B Instruct2B | Q4_K_M | 2.8 GB | ~128 tok/s ModelFit est. |
| LFM3 1.7B1.73B | Q4_K_M | 2.1 GB | ~199 tok/s ModelFit est. |
| Qwen2.5 1.5B Instruct1.5B | Q4_K_M | 2.5 GB | ~205 tok/s ModelFit est. |
| Llama 3.2 1B Instruct1B | Q4_K_M | 2.0 GB | ~228 tok/s ModelFit est. |
| Gemma 3 1B Instruct1B | Q8_0 | 2.0 GB | ~206 tok/s ModelFit est. |
| Qwen3.5 0.8B Instruct0.8B | Q4_K_M | 1.8 GB | ~242 tok/s ModelFit est. |
| Qwen2.5 0.5B Instruct0.5B | Q4_K_M | 1.8 GB | ~349 tok/s ModelFit est. |
| SmolLM2 360M0.36B | Q4_K_M | 1.5 GB | ~236 tok/s ModelFit est. |
Speeds come from ModelFit's estimates (CC BY 4.0) where it covers the model, otherwise from our bandwidth formula. Method
Too big for this machine
Llama 3.3 70B (44 GB), Llama 3.1 70B Instruct (44 GB), DeepSeek-R1 Distill Llama 70B (44 GB), Kolibri 1 (46 GB), Qwen3-Next 80B-A3B (53 GB), Nemotron 3 Super 120B-A12B (67 GB), gpt-oss-120b (68 GB), Llama 4 Scout (70 GB), Ling-3.0-flash (71 GB), Qwen3.5 122B-A10B Instruct (75 GB), Nemotron 3.5 Super VL 120B-A12B (76 GB), Devstral 2 123B (78 GB) and more.
Questions
What is the best AI model for the M3 Max (48 GB, 40-core GPU)?
Qwen3.6-35B-A3B at Q4_K_M: the largest model on our shortlist of current models that runs at 4-bit or better at a usable speed (about 109 tok/s, ModelFit est.).
How much memory can a model use on the M3 Max (48 GB, 40-core GPU)?
About 36 GB of the 48 GB.
How fast is the M3 Max (48 GB, 40-core GPU) for local LLMs?
Decode speed is limited by memory bandwidth: 400 GB/s. A model that reads 10 GB per token runs at under 40 tokens per second.