# M4 Max (16-core CPU / 40-core GPU)

> OnPrem World · https://onpremworld.com/en/hardware/apple-m4-max-40c-gpu · Last updated: 2026-10-14 · CC BY 4.0

Up to 128 GB of unified memory at 546 GB/s: a 128 GB MacBook Pro or Mac Studio runs 120B-class MoE models such as gpt-oss-120b.

## Specifications

|  |  |
|---|---|
| Memory | 48–128 GB LPDDR5X |
| Memory bandwidth | 546 GB/s |
| Launch | 2024-11 |
| Launch price | US$3,699 |
| Architecture | Apple M4 family (second-generation 3 nm; TSMC N3E) |
| Software | Metal / MLX |

## What it can run (128 GB, 8K)

Estimates, not benchmarks. See the methodology.

| Model | Parameters | Best quant | Total | tok/s (est.) |
|---|---|---|---|---|
| Qwen3.5-122B-A10B | 122B | Q5_K_M | 85.5 GiB | 50 |
| Mistral Small 4 119B (2603) | 119B | Q5_K_M | 83.4 GiB | 77 |
| gpt-oss-120b | 117B | Q6_K | 95.0 GiB | 84 |
| Llama 4 Scout 17B-16E Instruct | 109B | Q6_K | 89.8 GiB | 25 |
| GLM-4.5-Air | 106B | Q6_K | 87.3 GiB | 36 |
| Qwen3-Coder-Next (80B-A3B) | 80B | Q8_0 | 83.8 GiB | 111 |
| Llama 3.3 70B Instruct | 70B | Q8_0 | 75.7 GiB | 4.8 |
| Qwen3.5-35B-A3B | 35B | FP16 | 69.1 GiB | 59 |
| Qwen3.6-35B-A3B | 35B | FP16 | 69.1 GiB | 59 |
| Qwen3-32B | 32.8B | FP16 | 66.6 GiB | 5.4 |
| DeepSeek-R1-Distill-Qwen-32B | 32.76B | FP16 | 66.6 GiB | 5.4 |
| Gemma 4 31B IT | 30.7B | FP16 | 61.6 GiB | 5.8 |
| Qwen3-30B-A3B (2507) | 30.5B | FP16 | 60.9 GiB | 54 |
| Qwen3-Coder-30B-A3B-Instruct | 30.5B | FP16 | 60.9 GiB | 54 |
| GLM-4.7-Flash (30B-A3B) | 30B | FP16 | 59.6 GiB | 59 |
| Qwen3.5-27B | 27B | FP16 | 53.8 GiB | 6.6 |
| Qwen3.6-27B | 27B | FP16 | 53.8 GiB | 6.6 |
| Qwen3.8-27B | 27B | FP16 | 53.8 GiB | 6.6 |
| Gemma 3 27B IT | 27B | FP16 | 54.3 GiB | 6.6 |
| Gemma 4 26B A4B IT | 25.2B | FP16 | 50.1 GiB | 47 |
| Mistral Small 3.2 24B Instruct (2506) | 24B | FP16 | 48.7 GiB | 7.4 |
| Magistral Small 1.2 (2509) | 24B | FP16 | 48.7 GiB | 7.4 |
| Devstral Small 2 24B Instruct (2512) | 24B | FP16 | 48.7 GiB | 7.4 |
| gpt-oss-20b | 21B | FP16 | 41.8 GiB | 49 |
| Phi-4 (14B) | 14B | FP16 | 29.4 GiB | 13 |
| Gemma 3 12B IT | 12B | FP16 | 24.8 GiB | 15 |
| Gemma 4 12B Unified IT | 11.95B | FP16 | 24.2 GiB | 15 |
| Qwen3.5-9B | 9B | FP16 | 18.4 GiB | 20 |
| Qwen3-8B | 8.2B | FP16 | 17.7 GiB | 22 |
| DeepSeek-R1-0528-Qwen3-8B | 8.19B | FP16 | 17.6 GiB | 22 |
| Llama 3.1 8B Instruct | 8B | FP16 | 17.1 GiB | 22 |
| Gemma 4 E4B IT | 8B | FP16 | 16.4 GiB | 22 |
| Llama 3.2 3B Instruct | 3.21B | FP16 | 7.7 GiB | 55 |
| Kimi K2.6 | 1000B | Doesn't fit | – | – |
| DeepSeek-R1-0528 (671B) | 671B | Doesn't fit | – | – |
| DeepSeek-V4-Flash | 284B | Doesn't fit | – | – |
| Qwen3.8-Flash-Next | 180B | Doesn't fit | – | – |

## Prices

No prices recorded for this device yet.

## Frequently asked questions

### Which LLMs can the M4 Max (16-core CPU / 40-core GPU) run?

With 128 GB and an 8K context, the largest models that fit include: Qwen3.5-122B-A10B, Mistral Small 4 119B (2603), gpt-oss-120b, Llama 4 Scout 17B-16E Instruct, GLM-4.5-Air. See the full table below.

### How fast is the M4 Max (16-core CPU / 40-core GPU) for local LLMs?

Generation speed is mostly set by memory bandwidth (546 GB/s). For Llama 3.1 8B Instruct at FP16, expect roughly 22 tokens/s; bigger models are slower.

### Can the M4 Max (16-core CPU / 40-core GPU) run 70B models?

Yes. Llama 3.3 70B Instruct fits on the M4 Max (16-core CPU / 40-core GPU) (128 GB) at Q8_0 with an 8K context, using about 75.7 GiB, at an estimated 4.8 tokens/s.

## Sources

- https://support.apple.com/en-us/121553
- https://en.wikipedia.org/wiki/Apple_M4
- https://www.apple.com/newsroom/2024/10/new-macbook-pro-features-m4-family-of-chips-and-apple-intelligence/
- https://web.archive.org/web/20250107193915/https://www.apple.com/shop/buy-mac/macbook-pro/14-inch-space-black-standard-display-apple-m4-max-with-14-core-cpu-and-32-core-gpu-36gb-memory-1tb
- https://www.apple.com/newsroom/2024/10/apple-introduces-m4-pro-and-m4-max/