M1 Max (64 GB) — benchmark record

M1 Max (64 GB) local LLM benchmarks on Apple Silicon Mac with 10 published rows across 5 models. Peak published speed is 58.5 tok/s on Qwen3-Coder-30B-A3B. Largest published fit uses 22.0 GB. Published runtimes include llama.cpp, LM Studio (llama.cpp), LM Studio (MLX), MLX.

10Benchmark rows

5Models tested

58.5Fastest avg tok/s (Qwen3-Coder-30B-A3B)

0Lab benchmarks

Quick take

Best published speed here is 58.5 tok/s on Qwen3-Coder-30B-A3B at IQ4_XS. Largest published memory footprint is Nemotron-3-Nano-30B-A3B at Q4_K_XL, using 22.0 GB. Longest published context on this page is 8k. This page is evidence, not the full buying answer.

Based on 10 external benchmarks; no lab runs yet.

Published runtimes: llama.cpp, LM Studio (llama.cpp), LM Studio (MLX), MLX.

Need a recommendation? Use Buy Checking a target model? Use Fit Want a setup answer? Use Run Browse Macs by exact hardware Need the full audit trail? Use Bench

Current published coverage

Published model coverage includes Qwen3-Coder-30B-A3B, Qwen3.5-35B-A3B, Nemotron-3-Nano-30B-A3B, GLM-4.7-Flash plus 1 more published model. Published runtime coverage on this chip includes llama.cpp, LM Studio (llama.cpp), LM Studio (MLX), MLX. Fastest published row is 58.5 tok/s on Qwen3-Coder-30B-A3B at IQ4_XS. Largest published footprint on this page is 22.0 GB for Nemotron-3-Nano-30B-A3B. Longest published context is 8k.

Qwen3-Coder-30B-A3B Qwen3.5-35B-A3B Nemotron-3-Nano-30B-A3B GLM-4.7-Flash Qwen3.5-27B

Raw benchmark rows for M1 Max (64 GB)

Rows sorted by avg tok/s descending. Click source badge to see original measurement page.

Model	Quant	RAM req.	Context	Avg tok/s	Prompt tok/s	Runtime	Source
Qwen3-Coder-30B-A3B	IQ4_XS	16.1 GB	4k	58.5 tok/s	132.1 tok/s	llama.cpp	ref
Qwen3.5-35B-A3B	4bit	—	8k	57.6 tok/s	431.0 tok/s	MLX	ref
Qwen3.5-35B-A3B	4bit	—	—	57.0 tok/s	—	LM Studio (MLX)	ref
Nemotron-3-Nano-30B-A3B	Q4_K_XL	22.0 GB	4k	43.7 tok/s	136.9 tok/s	llama.cpp	ref
GLM-4.7-Flash	Q4_K_XL	17.0 GB	4k	36.8 tok/s	99.4 tok/s	llama.cpp	ref
Qwen3.5-35B-A3B	Q4_K - Medium	—	—	29.0 tok/s	—	LM Studio (llama.cpp)	ref
Qwen3.5-27B	4bit	—	8k	15.0 tok/s	67.0 tok/s	MLX	ref
Qwen3.5-27B	Q6_K	—	—	12.0 tok/s	—	llama.cpp	ref
Qwen3.5-27B	Q4_K - Medium	—	—	11.5 tok/s	—	llama.cpp	ref
Qwen3.5-27B	Q8_0	—	—	10.5 tok/s	—	llama.cpp	ref

Models tested on this chip

Qwen3-Coder-30B-A3B Qwen3.5-35B-A3B Nemotron-3-Nano-30B-A3B GLM-4.7-Flash Qwen3.5-27B

Next step

Use Rankings when you need the best overall answer for a Mac, and keep this page open when you need the evidence behind that recommendation.

If you are comparing local Apple Silicon against rented cloud hardware, use AI Data Center Index for current GPU rental context.

Data

benchmarks.json — full dataset · chips.json — chip summaries · benchmarks.csv — CSV export

Data from in-house lab measurements plus community-published benchmarks. See all chip families →