Local LLM Performance on Apple Silicon

Detailed per-chip benchmark pages for 73 model+chip combinations. Find out exactly how fast any model runs on your Mac — from M1 with 8 GB RAM to M5 Max with 128 GB.

These pages answer how fast is this model on that chip. For the inverse — what should I run on the Mac I already own — go to Best local LLM for every Mac →
Shopping for the machine instead? The Mac Advisor turns a budget and a use case into one config → · Every Mac compared →

Gemma 4 E2B

2.3B • up to 261 tok/s

1 chips

M5 Max 261 tok/s 128GB

Qwen 3.5 4B

4B • up to 161 tok/s

1 chips

M5 Max 161 tok/s 64GB

Gemma 4 E4B

4B • up to 161 tok/s

4 chips

M5 Max 161 tok/s 128GB M5 Pro 84 tok/s 24GB M4 Pro 84 tok/s 24GB M6 55 tok/s 16GB

Qwen 3 8B

8B • up to 91 tok/s

4 chips

M5 Max 91 tok/s 128GB M4 Pro 45 tok/s 24GB M2 17 tok/s 16GB M1 12 tok/s 16GB

Llama 3.1 8B

8B • up to 91 tok/s

3 chips

M5 Max 91 tok/s 128GB M4 20 tok/s 16GB M1 12 tok/s 16GB

Mistral Small 4

119B • up to 86 tok/s

1 chips

M5 Ultra 86 tok/s 96GB

Qwen 3.5 9B

9B • up to 82 tok/s

6 chips

M5 Max 82 tok/s 64GB M4 Pro 40 tok/s 24GB M6 26 tok/s 16GB M4 18 tok/s 16GB M3 15 tok/s 16GB M1 10 tok/s 16GB

LFM2.5-2.6B

2.6B • up to 81 tok/s

1 chips

M6 81 tok/s 16GB

DeepSeek V4 Flash

284B • up to 80 tok/s

4 chips

M5 Ultra 80 tok/s 256GB M5 Max 39 tok/s 128GB M3 Ultra 30 tok/s 256GB M3 Max 27 tok/s 128GB

Qwen3-Coder-Next

80B • up to 72 tok/s

2 chips

M5 Ultra 72 tok/s 96GB M5 Max 35 tok/s 64GB

Gemma 3 12B

12B • up to 64 tok/s

3 chips

M5 Max 64 tok/s 64GB M3 12 tok/s 16GB M2 12 tok/s 16GB

GLM-4.5-Air

106B • up to 61 tok/s

1 chips

M5 Ultra 61 tok/s 96GB

Qwen3.8-27B

27.8B • up to 57 tok/s

2 chips

M5 Ultra 57 tok/s 96GB M6 8 tok/s 24GB

Laguna S 2.1

118B • up to 53 tok/s

1 chips

M5 Ultra 53 tok/s 96GB

KAT-Coder-V2.5

35B-A3B • up to 52 tok/s

2 chips

M5 Max 52 tok/s 64GB M6 15 tok/s 32GB

Gemma 4 26B-A4B

26B • up to 50 tok/s

3 chips

M5 Max 50 tok/s 128GB M4 Pro 28 tok/s 24GB M6 14 tok/s 24GB

Gemma 3 4B

4B • up to 49 tok/s

1 chips

M3 Pro 49 tok/s 18GB

Inkling-Small

276B • up to 45 tok/s

1 chips

M5 Ultra 45 tok/s 256GB

Qwen 3.6-35B-A3B

35B • up to 42 tok/s

2 chips

M4 Max 42 tok/s 48GB M6 14 tok/s 32GB

Qwen 3.6-27B

27B • up to 39 tok/s

6 chips

M1 Ultra 39 tok/s 64GB M5 Max 30 tok/s 64GB M4 Max 27 tok/s 64GB M2 Max 20 tok/s 32GB M5 Pro 14 tok/s 24GB M6 9 tok/s 24GB

Qwen3-235B-A22B

235B • up to 37 tok/s

1 chips

M5 Ultra 37 tok/s 256GB

DeepSeek R1 8B

8B • up to 29 tok/s

1 chips

M6 29 tok/s 16GB

Gemma 3 27B

27B • up to 27 tok/s

2 chips

M4 Max 27 tok/s 48GB M4 Pro 14 tok/s 24GB

Muse Glimmer 30B

30B • up to 27 tok/s

2 chips

M5 Max 27 tok/s 64GB M6 8 tok/s 32GB

Gemma 4 31B

31B • up to 26 tok/s

2 chips

M5 Max 26 tok/s 128GB M4 Pro 12 tok/s 24GB

Qwen 3 14B

14B • up to 26 tok/s

2 chips

M4 Pro 26 tok/s 24GB M3 10 tok/s 16GB

DeepSeek R1 70B

70B • up to 24 tok/s

2 chips

M5 Ultra 24 tok/s 96GB M5 Max 12 tok/s 128GB

GLM-4.7-Flash

31B • up to 24 tok/s

1 chips

M4 Max 24 tok/s 48GB

Llama 3.3 70B

70B • up to 24 tok/s

1 chips

M5 Ultra 24 tok/s 96GB

Qwen 3 32B

32B • up to 23 tok/s

1 chips

M4 Max 23 tok/s 64GB

Qwen 2.5 72B

72B • up to 23 tok/s

1 chips

M5 Ultra 23 tok/s 96GB

Phi-4 14B

14B • up to 17 tok/s

2 chips

M6 17 tok/s 16GB M4 12 tok/s 16GB

Nemotron 3.5 Lightning

30B • up to 17 tok/s

1 chips

M6 17 tok/s 32GB

GLM 5.2

753B-A40B • up to 15 tok/s

1 chips

M3 Ultra 15 tok/s 512GB

GPT-oss 120B

117B • up to 14 tok/s

3 chips

M5 Ultra 14 tok/s 96GB M4 Ultra 13 tok/s 192GB M5 Max 7 tok/s 128GB

Llama 3.1 405B

405B • up to 4 tok/s

1 chips

M5 Ultra 4 tok/s 512GB