RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?OllamaStart hereAPI pricesVRAM calculatorNew

Can Kimi K2.5 run on the Mac M3 Ultra 512GB?

Yes. At 8K context the best fit is Q2_K, using about 360 GB of the 384 GB usable on the Mac M3 Ultra 512GB. Estimated speed ceiling: ~66 tokens/sec.

+ 0.6 GB with image input. The figures on this page are for text only. With image input enabled, the vision projector (mmproj) file loads alongside the model: Q2_K at 8K context then needs about 360 GB (tight). llama.cpp's -hf downloads and loads it automatically unless you add --no-mmproj. Calculator with image input on.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K)360 GB Tight360 GB Tight361 GB Tight368 GB Tight66
3-bit (Q3_K_M)468 GB No469 GB No470 GB No477 GB No—
4-bit (Q4_K_M)586 GB No586 GB No587 GB No594 GB No—
5-bit (Q5_K_M)682 GB No683 GB No684 GB No691 GB No—
6-bit (Q6_K)785 GB No785 GB No787 GB No794 GB No—
8-bit (Q8_0)1017 GB No1017 GB No1019 GB No1025 GB No—
16-bit (F16/BF16)1914 GB No1914 GB No1916 GB No1922 GB No—

• real GGUF file size. Speed is a bandwidth ceiling (819 GB/s), not a benchmark. Methodology.

What fits instead

"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.

More

Kimi K2.5 on every GPU and Mac · Every model that runs on the Mac M3 Ultra 512GB · Apple Silicon guide · Try other settings in the calculator

Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs

Data: Hugging Face Hub (moonshotai/Kimi-K2.5), Apple specs. Model data last checked 2026-11-18; refreshed hourly.