Can I run it?
Can I run gemma 4 26B A4B it on Apple M3 Pro?
Tight fit
Tight fit fully in unified memory on your Apple M3 Pro at Q3_K_M, 4K context, about 13–18 tok/s. Weights 11.9 GB plus 0.35 GB of KV cache — 12.6 GB of 13.0 GB usable.
Best quantQ3_K_M
Working context4K
Max that fits4K
Speed13–18 tok/sEstimate from memory bandwidth, not a benchmark. Method.
Weights11.9 GB
KV cache0.35 GB
Total12.6 GB
Usable13.0 GB
gemma 4 26B A4B it details · Apple M3 Pro page · How this estimate is made
Run commands
llama.cpp
llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:Q3_K_M -c 4096 -ngl 30
Ollama
ollama run gemma
LM Studio
lms get unsloth/gemma-4-26B-A4B-it-GGUF # Then load quant Q3_K_M
Assumptions
- Estimate from memory bandwidth, not a benchmark. Decode speed uses 55–75% of peak bandwidth.
- KV cache is FP16. Engine overhead assumes llama.cpp (0.3 GB + 1% of the weights). Ollama is closer to 0.9 GB + 3%, LM Studio closer to 1 GB + 3%.
- Q3_K_M size is the real GGUF file size from Hugging Face.
- Unified memory uses 75% for the GPU by default, and always leaves about 5 GB for the OS.
- Speed is quoted at 4K context, not at the maximum context that still fits.
Larger models that also fit
- Can I run Qwen3.8 27B on Apple M3 Pro? — Tight fit, UD-IQ3_S, 6.6–8.9 tok/s (estimate)
- Can I run Qwen3.6 35B A3B on Apple M3 Pro? — Tight fit, UD-Q2_K_XL, 16–22 tok/s (estimate)
- Can I run Qwen3.5 35B A3B on Apple M3 Pro? — Tight fit, UD-IQ3_XXS, 15–21 tok/s (estimate)
Updated 2026-10-24. Profile on this page is the device default.