What LLMs Can the Apple M4 Max 128GB Run?
The Apple M4 Max 128GB has 128 GB of unified memory (about 96.0 GB usable by the GPU) and 546 GB/s of bandwidth. It runs 50 of 58 models fully at Q4_K_M with 8K context, up to Qwen 3.5 122B-A10B, and generates about 64 tokens/s on an 8B model.
51 of 58 models fit
Apple M4 Max 128GB (96.0 GB GPU-usable) at Q4_K_M with 8K context. Largest that fits fully: Qwen 3.5 122B-A10B.
| Model | Needs | Status | Speed | Best quant that fits |
|---|---|---|---|---|
| Qwen 3.5 122B-A10B | 70.2 GB | Fits | ~44 tok/s | Q5_K_M |
| Mistral Small 4 119B | 68.2 GB | Fits | ~67 tok/s | Q5_K_M |
| gpt-oss-120b | 67.2 GB | Fits | ~78 tok/s | MXFP4 (official) |
| Llama 4 Scout | 64.0 GB | Fits | ~28 tok/s | Q6_K |
| GLM-4.5-Air | 62.1 GB | Fits | ~37 tok/s | Q6_K |
| Qwen3-Next 80B-A3B | 46.3 GB | Fits | ~96 tok/s | Q8_0 |
| Qwen 2.5 72B | 44.5 GB | Fits | ~7.8 tok/s | Q8_0 |
| Llama 3.3 70B | 43.2 GB | Fits | ~8.0 tok/s | Q8_0 |
| DeepSeek R1 Distill Llama 70B | 43.2 GB | Fits | ~8.0 tok/s | Q8_0 |
| Mixtral 8x7B | 28.0 GB | Fits | ~37 tok/s | FP16 / BF16 |
| Qwen 2.5 32B | 21.4 GB | Fits | ~17 tok/s | FP16 / BF16 |
| Qwen 2.5 Coder 32B | 21.4 GB | Fits | ~17 tok/s | FP16 / BF16 |
| DeepSeek R1 Distill Qwen 32B | 21.4 GB | Fits | ~17 tok/s | FP16 / BF16 |
| Qwen 3 32B | 21.4 GB | Fits | ~17 tok/s | FP16 / BF16 |
| Qwen 3.5 35B-A3B | 21.0 GB | Fits | ~105 tok/s | FP16 / BF16 |
| Qwen 3.6 35B-A3B | 21.0 GB | Fits | ~105 tok/s | FP16 / BF16 |
| Gemma 4 31B | 19.8 GB | Fits | ~18 tok/s | FP16 / BF16 |
| Qwen 3 30B-A3B | 18.8 GB | Fits | ~86 tok/s | FP16 / BF16 |
| Qwen3-Coder 30B-A3B | 18.8 GB | Fits | ~86 tok/s | FP16 / BF16 |
| Gemma 3 27B | 17.6 GB | Fits | ~20 tok/s | FP16 / BF16 |
| Qwen 3.5 27B | 16.9 GB | Fits | ~21 tok/s | FP16 / BF16 |
| Qwen 3.6 27B | 16.9 GB | Fits | ~21 tok/s | FP16 / BF16 |
| Qwen 3.8 27B | 16.9 GB | Fits | ~21 tok/s | FP16 / BF16 |
| Mistral Small 3.1 24B | 15.6 GB | Fits | ~23 tok/s | FP16 / BF16 |
| Gemma 4 26B-A4B | 15.6 GB | Fits | ~97 tok/s | FP16 / BF16 |
| gpt-oss-20b | 12.9 GB | Fits | ~113 tok/s | MXFP4 (official) |
| Qwen 2.5 14B | 10.7 GB | Fits | ~36 tok/s | FP16 / BF16 |
| DeepSeek R1 Distill Qwen 14B | 10.7 GB | Fits | ~36 tok/s | FP16 / BF16 |
| Phi-4 14B | 10.6 GB | Fits | ~36 tok/s | FP16 / BF16 |
| Qwen 3 14B | 10.5 GB | Fits | ~36 tok/s | FP16 / BF16 |
| Ministral 3 14B | 9.7 GB | Fits | ~39 tok/s | FP16 / BF16 |
| Mistral Nemo 12B | 9.0 GB | Fits | ~43 tok/s | FP16 / BF16 |
| Gemma 3 12B | 8.8 GB | Fits | ~43 tok/s | FP16 / BF16 |
| Ministral 3 8B | 6.7 GB | Fits | ~61 tok/s | FP16 / BF16 |
| Qwen 3 8B | 6.6 GB | Fits | ~62 tok/s | FP16 / BF16 |
| Qwen 3.5 9B | 6.4 GB | Fits | ~62 tok/s | FP16 / BF16 |
| Llama 3.1 8B | 6.3 GB | Fits | ~64 tok/s | FP16 / BF16 |
| DeepSeek R1 Distill Llama 8B | 6.3 GB | Fits | ~64 tok/s | FP16 / BF16 |
| Mistral 7B v0.3 | 5.7 GB | Fits | ~70 tok/s | FP16 / BF16 |
| Gemma 4 E4B | 5.7 GB | Fits | ~70 tok/s | FP16 / BF16 |
| Qwen 2.5 7B | 5.6 GB | Fits | ~72 tok/s | FP16 / BF16 |
| DeepSeek R1 Distill Qwen 7B | 5.6 GB | Fits | ~72 tok/s | FP16 / BF16 |
| Qwen 3 4B | 4.2 GB | Fits | ~112 tok/s | FP16 / BF16 |
| Phi-4 Mini | 4.1 GB | Fits | ~119 tok/s | FP16 / BF16 |
| Gemma 4 E2B | 4.0 GB | Fits | ~109 tok/s | FP16 / BF16 |
| Gemma 3 4B | 3.7 GB | Fits | ~121 tok/s | FP16 / BF16 |
| Qwen 3.5 4B | 3.7 GB | Fits | ~125 tok/s | FP16 / BF16 |
| Ministral 3 3B | 3.5 GB | Fits | ~136 tok/s | FP16 / BF16 |
| Llama 3.2 3B | 3.5 GB | Fits | ~141 tok/s | FP16 / BF16 |
| Llama 3.2 1B | 1.7 GB | Fits | ~376 tok/s | FP16 / BF16 |
| Gemma 3 1B | 1.6 GB | Fits | ~488 tok/s | FP16 / BF16 |
| Kimi K2.6 | 566 GB | Too small | – | – |
| DeepSeek R1 671B | 380 GB | Too small | – | – |
| Llama 3.1 405B | 234 GB | Too small | – | – |
| Qwen 3.5 397B-A17B | 226 GB | Too small | – | – |
| DeepSeek V4 Flash | 161 GB | Too small | – | – |
| Qwen 3 235B-A22B | 135 GB | Too small | – | Q2_K |
| MiniMax-M2 | 133 GB | Too small | – | Q2_K |
Apple M4 Max 128GB for local AI
The largest laptop memory pool. About 96 GB is GPU-usable by default (more with iogpu.wired_limit_mb), which runs 70B models at Q8_0 or gpt-oss-120b natively.
Setup tips: raise the GPU memory limit with sudo sysctl iogpu.wired_limit_mb=114688 to give models about 112 GB. Prefer MLX builds in LM Studio and mixture-of-experts models, which make the best use of its bandwidth. Keep the laptop plugged in for long runs.
Among the 5 devices we track with about 96 GB of usable memory, the Apple M4 Max 128GB ranks 3 of 5 for text-generation speed, because generation speed follows memory bandwidth. Others in the same class: RTX PRO 6000 Blackwell 96GB, Apple M5 Max 128GB, Apple M3 Max 128GB, Ryzen AI Max+ 395 128GB. They all hold the same models; the faster ones simply answer sooner.
Our picks for the Apple M4 Max 128GB
The model we would run first for each job, chosen from our recommendations by what fits fully with an 8K context. "Best quant" is the highest quality that fits; "max context" is at Q4_K_M or the native format.
| Task | Model | Best quant | Max context | Speed |
|---|---|---|---|---|
| Best all-rounder | Qwen 3.5 122B-A10B | Q5_K_M | 256K | ~44 tok/s |
| Coding | GLM-4.5-Air | Q6_K | 128K | ~37 tok/s |
| Fastest replies | Qwen 3.5 122B-A10B | Q5_K_M | 256K | ~44 tok/s |
| Long documents (64K+) | Qwen 3.5 122B-A10B | Q5_K_M | 256K | ~44 tok/s |
| Images in, text out | Gemma 4 31B | FP16 / BF16 | 256K | ~18 tok/s |
| Image generation | FLUX.2 [dev] | FP8 | – | – |
| Spec | Value |
|---|---|
| Memory | 128 GB unified (96.0 GB GPU-usable by default) |
| Memory bandwidth | 546 GB/s |
| Architecture | Apple Silicon |
| Released | 2024-10 |
| Llama 3.1 8B Q4_K_M speed | ~64 tok/s |
| Largest model that fits | Qwen 3.5 122B-A10B |
Image and video models on the Apple M4 Max 128GB
Peak VRAM with model swapping (ComfyUI default on smaller cards).
| Model | FP8 | FP8 fit | GGUF Q4 | Q4 fit |
|---|---|---|---|---|
| Stable Diffusion 1.5 | 1.8 GB | Fits | 1.7 GB | Fits |
| SDXL 1.0 | 4.1 GB | Fits | 4.1 GB | Fits |
| Stable Diffusion 3.5 Medium | 6.4 GB | Fits | 6.4 GB | Fits |
| Stable Diffusion 3.5 Large | 8.6 GB | Fits | 6.4 GB | Fits |
| FLUX.1 [dev] | 12.1 GB | Fits | 7.3 GB | Fits |
| FLUX.1 [schnell] | 12.1 GB | Fits | 7.3 GB | Fits |
| Qwen-Image | 20.4 GB | Fits | 12.3 GB | Fits |
| Wan 2.2 A14B (T2V / I2V) | 18.8 GB | Fits | 13.2 GB | Fits |
| Wan 2.2 TI2V 5B | 8.0 GB | Fits | 6.5 GB | Fits |
| HunyuanVideo | 18.3 GB | Fits | 13.1 GB | Fits |
| FLUX.1 Kontext [dev] | 12.3 GB | Fits | 7.5 GB | Fits |
| FLUX.2 [dev] | 31.0 GB | Fits | 23.6 GB | Fits |
| Z-Image Turbo | 6.8 GB | Fits | 4.9 GB | Fits |
| Qwen-Image-Edit 2509 | 20.5 GB | Fits | 12.4 GB | Fits |
| LTX-2 | 19.2 GB | Fits | 12.6 GB | Fits |
Upgrade path
The next step up is the Apple M2 Ultra 192GB with 192 GB and 47% more bandwidth.
Comparisons
- Apple M4 Max 128GB vs RTX 5090 32GB
- Apple M4 Max 128GB vs Apple M4 Pro 48GB
- Apple M3 Ultra 512GB vs Apple M4 Max 128GB
- Apple M4 Max 128GB vs Apple M5 Max 128GB
- Apple M4 Max 128GB vs NVIDIA DGX Spark 128GB
Apple M4 Max 128GB for LLMs: common questions
Can the Apple M4 Max 128GB run Llama 3.3 70B?
Yes. Llama 3.3 70B needs about 43.2 GB at Q4_K_M with 8K context, and the best quality that fits is Q8_0.
Can the Apple M4 Max 128GB run Qwen 3 32B?
Yes. Qwen 3 32B needs about 21.4 GB at Q4_K_M with 8K context, and the best quality that fits is FP16 / BF16.
Can the Apple M4 Max 128GB run Qwen 3 14B?
Yes. Qwen 3 14B needs about 10.5 GB at Q4_K_M with 8K context, and the best quality that fits is FP16 / BF16.
How fast is the Apple M4 Max 128GB for LLMs?
About 64 tokens per second for an 8B model at Q4_K_M, based on its 546 GB/s memory bandwidth. Larger models scale down roughly in proportion to their size.
How much of the 128 GB can the GPU use?
About 96.0 GB by default. You can raise it with sudo sysctl iogpu.wired_limit_mb=<MB> (resets on reboot), leaving at least 8 GB for macOS.