M4 Pro (24 GB) — benchmark record

M4 Pro (24 GB) local LLM benchmarks on Apple Silicon Mac with 19 published rows across 17 models. Peak published speed is 118.0 tok/s on Qwen 3 4B. Published runtimes include LM Studio, MLX, Ollama.

19Benchmark rows

17Models tested

118.0Fastest avg tok/s (Qwen 3 4B)

0Lab benchmarks

Quick take

Best published speed here is 118.0 tok/s on Qwen 3 4B at Q4_K - Medium. This page is evidence, not the full buying answer.

Based on 19 external benchmarks; no lab runs yet.

Published runtimes: LM Studio, MLX, Ollama.

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 Qwen 3 4B, Phi-4 Mini Instruct 3.8B, Mistral 7B v0.3, Gemma 4 E2B plus 13 more published models. Published runtime coverage on this chip includes LM Studio, MLX, Ollama. Fastest published row is 118.0 tok/s on Qwen 3 4B at Q4_K - Medium.

Qwen 3 4B Phi-4 Mini Instruct 3.8B Mistral 7B v0.3 Gemma 4 E2B Gemma 3 4B Qwen3.5-9B

Other published M4 Pro variants: 32 GB · 48 GB · 64 GB

Macs shipping with the M4 Pro (24 GB)

Mac Mini M4 Pro 24GB Mac Mini M4 Pro 48GB MacBook Pro M4 Pro 24GB 14-inch MacBook Pro M4 Pro 48GB 14-inch MacBook Pro M4 Pro 24GB 16-inch MacBook Pro M4 Pro 48GB 16-inch

This chip is part of a family. Compare all M4 Pro RAM variants →

Raw benchmark rows for M4 Pro (24 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
Qwen 3 4B	Q4_K - Medium	—	—	118.0 tok/s	—	MLX	ref
Phi-4 Mini Instruct 3.8B	Q4_K - Medium	—	—	108.0 tok/s	—	Ollama	ref
Mistral 7B v0.3	Q4_K - Medium	—	—	98.0 tok/s	—	MLX	ref
Gemma 4 E2B	Q4_K - Medium	—	—	95.0 tok/s	—	Ollama	ref
Gemma 3 4B	Q8_0	—	—	95.0 tok/s	—	MLX	ref
Qwen3.5-9B	Q4_K - Medium	—	—	92.0 tok/s	—	MLX	ref
Qwen 3 8B	Q4_K - Medium	—	—	82.0 tok/s	—	Ollama	ref
Gemma 4 E4B	Q4_K - Medium	—	—	78.0 tok/s	—	MLX	ref
Qwen 3 8B	Q8_0	—	—	68.0 tok/s	—	MLX	ref
Mistral 7B v0.3	Q8_0	—	—	65.0 tok/s	—	Ollama	ref
Gemma 3 12B	Q4_K - Medium	—	—	52.0 tok/s	—	MLX	ref
Ministral 3 14B	Q4_K - Medium	—	—	40.0 tok/s	—	Ollama	ref
Qwen 3 14B	Q4_K - Medium	—	—	38.0 tok/s	—	LM Studio	ref
Qwen 3 30B-A3B	Q4_K - Medium	—	—	35.0 tok/s	—	MLX	ref
Qwen3.6-35B-A3B	Q4_K - Medium	—	—	32.0 tok/s	—	Ollama	ref
Gemma 4 26B-A4B	Q4_K - Medium	—	—	28.0 tok/s	—	Ollama	ref
Gemma 3 27B	Q4_K - Medium	—	—	25.0 tok/s	—	LM Studio	ref
Nemotron Cascade 2 30B-A3B	Q4_K - Medium	—	—	22.0 tok/s	—	Ollama	ref
Gemma 4 31B	Q4_K - Medium	—	—	14.0 tok/s	—	Ollama	ref

Models tested on this chip

Qwen 3 4B Phi-4 Mini Instruct 3.8B Mistral 7B v0.3 Gemma 4 E2B Gemma 3 4B Qwen3.5-9B Qwen 3 8B Gemma 4 E4B Gemma 3 12B Ministral 3 14B Qwen 3 14B Qwen 3 30B-A3B Qwen3.6-35B-A3B Gemma 4 26B-A4B Gemma 3 27B Nemotron Cascade 2 30B-A3B Gemma 4 31B

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 →