GPUs & Macs › Apple M2 Pro

Apple M2 Pro: local LLM speed

Llama 2 7B (Q4_0) generates 21.0 tokens/s on the Apple M2 Pro (Vulkan), and processes prompts at 59 tokens/s.

Measured results

ModelGenerate
tok/s		Prompt
tok/s		QuantBackendTestSource
Llama 2 7B21.059Q4_0Vulkan FA ontg128llama.cpp #10879 details
Llama 2 7B20.963Q4_0Vulkan FA offtg128llama.cpp #10879 details

What fits on the Apple M2 Pro

Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 21.0 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.

With 16 GB unified memory

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BUD-IQ2_XXS7.3 GB11.0 estimate
gemma-4-12B-itQ6_K9.8 GB8.2 estimate
Qwen3.5-9BQ8_09.5 GB8.4 estimate
Qwen3.5-4BQ8_04.5 GB17.9 estimate
Qwen3.6-35B-A3BUD-IQ1_M10.0 GB46.6 estimate
gemma-4-26B-A4B-itUD-IQ2_M10.0 GB26.0 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB8.4 estimate
gemma-4-E4B-itQ8_08.2 GB9.8 estimate

With 36 GB unified memory

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BUD-Q6_K22.0 GB3.6 estimate
gemma-4-12B-itQ8_012.7 GB6.3 estimate
Qwen3.5-9BQ8_09.5 GB8.4 estimate
Qwen3.5-4BQ8_04.5 GB17.9 estimate
Qwen3.6-35B-A3BUD-Q5_K_S24.9 GB18.8 estimate
gemma-4-26B-A4B-itUD-Q6_K23.2 GB11.3 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB8.4 estimate
gemma-4-E4B-itQ8_08.2 GB9.8 estimate

With 64 GB unified memory

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BQ8_029.0 GB2.8 estimate
gemma-4-12B-itQ8_012.7 GB6.3 estimate
Qwen3.5-9BQ8_09.5 GB8.4 estimate
Qwen3.5-4BQ8_04.5 GB17.9 estimate
Qwen3.6-35B-A3BQ8_036.9 GB12.7 estimate
gemma-4-26B-A4B-itQ8_026.9 GB9.7 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB8.4 estimate
gemma-4-E4B-itQ8_08.2 GB9.8 estimate

With 128 GB unified memory

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BQ8_029.0 GB2.8 estimate
gemma-4-12B-itQ8_012.7 GB6.3 estimate
Qwen3.5-9BQ8_09.5 GB8.4 estimate
Qwen3.5-4BQ8_04.5 GB17.9 estimate
Qwen3.6-35B-A3BQ8_036.9 GB12.7 estimate
Qwen3.8-Flash-NextUD-IQ3_XXS82.0 GB1.0 estimate
gemma-4-26B-A4B-itQ8_026.9 GB9.7 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB8.4 estimate

Open in the picker

Similar hardware

Apple M1 (7-core GPU) · Apple M1 (8-core GPU) · Apple M1 Pro (14-core GPU) · Apple M1 Pro (16-core GPU) · Apple M1 Max (24-core GPU) · Apple M1 Max (32-core GPU) · Apple M1 Ultra (48-core GPU) · Apple M1 Ultra (64-core GPU)