Can a Mac Studio M5 Max, 36GB run local LLMs?
Yes — 27 of the 39 open models on this site fit in its 27 GB of usable memory, the strongest being Qwen3.8 27B. Whether that saves you money is a different question, and the answer is usually no.
Price$2,499
Memory36 GB, about 27 GB of it addressable by the GPU at 460 GB/s
Best model it runsQwen3.8 27B — Sonnet-class, 19 tok/s
Pay-back against the APIPays back in 32 years at 500k tokens a day
Run the numbers on this machine
What it runs
| Model | Speed | Class | Good at | Memory |
|---|---|---|---|---|
| Qwen3.8 27BQ4_K_M | 19 tok/s | Sonnet-class | 19 GB | |
| Qwen3.6 27BQ4_K_M | 18 tok/s | Haiku-class | 19 GB | |
| Qwen3.6 35B-A3BQ4_K_M | 55 tok/s | Haiku-class | 23 GB | |
| Muse Glimmer 30BQ4_K_M | 19 tok/s | Haiku-class | 18 GB | |
| Gemma 4 26B-A4BQ4_K_M | 40 tok/s | Haiku-class | 18 GB | |
| Gemma 4 31B itQ4_K_M | 13 tok/s | Haiku-class | 26 GB | |
| Granite 4.2 30BQ4_K_M | 13 tok/s | Haiku-class | 26 GB | |
| GLM-4.7-FlashQ4_K_M | 39 tok/s | Haiku-class | 20 GB | |
| Gemma 4 12BQ4_K_M | 43 tok/s | Below every hosted tier | 8.0 GB | |
| Qwen3.5 9BQ4_K_M | 51 tok/s | Below every hosted tier | 6.8 GB | |
| Nemotron 3.5 Lightning 30B-A3BQ4_K_M | 53 tok/s | Below every hosted tier | 26 GB | |
| MiniCPM5 2BQ4_K_M | 116 tok/s | Below every hosted tier | 3.0 GB |
15 more fit; the calculator lists them all.
The specifics
- Chip
- M5 Max — 18-core CPU / 32-core GPU
- Memory bandwidth
- 460 GB/s
- Usable by the GPU
- 27 GB — macOS lets the GPU wire roughly 75% of unified memory by default (llama.cpp discussion #2182: 66.7% at 32 GiB or less, 75% above). Raise it with `sudo sysctl iogpu.wired_limit_mb`. Memory is treated as GB throughout, which is slightly conservative.
- Power under load
- 145 W (stand in) — Apple has not published power figures for the 2026 machines yet (support articles 103253 / 102027 stop at the 2024/2025 models). Stand-in: Apple's published maximum for the previous chip (M4 Max Mac Studio, 145 W), which is a CPU+GPU stress figure, so it likely overstates inference draw.
- Availability
- Pre-order; ships 2026-09-22.
- Sources
- source 1, source 2, source 3