M1 Ultra (64 GB) — benchmark record

M1 Ultra (64 GB) local LLM benchmarks on Apple Silicon Mac with 5 published rows across 2 models. Peak published speed is 29.7 tok/s on Devstral Small 2 24B. Published runtimes include llama.cpp, LM Studio, MLX.

5Benchmark rows

2Models tested

29.7Fastest avg tok/s (Devstral Small 2 24B)

0Lab benchmarks

Quick take

Best published speed here is 29.7 tok/s on Devstral Small 2 24B at 4bit. Longest published context on this page is 4k. This page is evidence, not the full buying answer.

Based on 5 external benchmarks; no lab runs yet.

Published runtimes: llama.cpp, LM Studio, 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 Devstral Small 2 24B, Llama 3.3 70B. Published runtime coverage on this chip includes llama.cpp, LM Studio, MLX. Fastest published row is 29.7 tok/s on Devstral Small 2 24B at 4bit. Longest published context is 4k.

Devstral Small 2 24B Llama 3.3 70B

Raw benchmark rows for M1 Ultra (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
Devstral Small 2 24B	4bit	—	—	29.7 tok/s	—	MLX	ref
Devstral Small 2 24B	Q4_K - Medium	—	—	25.3 tok/s	—	llama.cpp	ref
Devstral Small 2 24B	Q8	—	—	23.4 tok/s	—	llama.cpp	ref
Devstral Small 2 24B	8bit	—	—	22.3 tok/s	—	MLX	ref
Llama 3.3 70B	4bit	—	4k	12.6 tok/s	—	LM Studio	ref

Models tested on this chip

Devstral Small 2 24B Llama 3.3 70B

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 →