Skip to content

Can I run DeepSeek, Kimi or GLM locally?

Yes, if you have enough memory. DeepSeek V3.1 needs about 378 GB at 4-bit, which runs comfortably on three 256 GB Mac Studios. Kimi K2 Thinking needs about 594 GB, or four. DeepSeek V4 Pro ships at about 893 GB, or six. With exo, four Mac Studios ran DeepSeek V3.1 at 32.5 tokens per second.

These are mixture-of-experts models: hundreds of billions of parameters in total, but only 32 to 49 billion active per token, which is why they run at usable speeds on Macs.

Smaller models fit on one machine: Qwen3.8-27B runs on a Mac with 32 GB, and gpt-oss-120b on one with 96 to 128 GB.

Check each model’s license on Hugging Face before commercial use.

Sources

  1. Jeff Geerling: 1.5 TB of VRAM on Mac Studio, RDMA over Thunderbolt 5
  2. Hugging Face: deepseek-ai/DeepSeek-V3.1
  3. Hugging Face: moonshotai/Kimi-K2-Thinking
  4. Hugging Face: zai-org/GLM-5.3
  5. Hugging Face: deepseek-ai/DeepSeek-V4-Pro-0813
  6. exo on GitHub (README)

Updated

Planning local AI for your team?

Tell us about your models, users and hardware. Exo Labs helps teams size, build and run private AI clusters.

Talk to Exo Labs →Download exoCheck your hardware on local.ai ↗