@antirez: I didn't expect DeepSeek v4 PRO (not Flash) to run well on the Mac Studio M3 Ultra with 512GB of RAM. This is 2 bit qua…
Summary
Antirez reports that DeepSeek v4 PRO runs well on a Mac Studio M3 Ultra with 512GB RAM using 2-bit quantization, achieving 130 t/s prefill and 13 t/s generation.
View Cached Full Text
Cached at: 05/17/26, 11:32 AM
I didn’t expect DeepSeek v4 PRO (not Flash) to run well on the Mac Studio M3 Ultra with 512GB of RAM. This is 2 bit quantized with the same DwarfStar recipe used for Flash. 433GB GGUF file. 130 t/s prefill, 13 t/s generation. Prefill in the video is low because small prompt. https://t.co/ciyx0XCSh7
Similar Articles
You can run Deepseek 4 flash on mac (M3 Max, 96gb)
A guide on running DeepSeek 4 flash on a Mac M3 Max with 96GB RAM using Antirez's ds4 engine and SSD streaming, achieving ~12 tokens/second inference speed.
@antirez: DeepSeek v4 PRO running via SSD streaming on my 128GB MacBook m5 max. 1.6 trillion parameters.
DeepSeek v4 PRO, a 1.6 trillion parameter model, is running via SSD streaming on a 128GB MacBook m5 max, demonstrating local inference of a massive model.
@Snixtp: DeepSeek V4 Flash on a single RTX Pro 6000?
DeepSeek V4 Flash GGUF quantizations have been released by antirez, enabling the model to run on single GPUs like the RTX Pro 6000 and Macs with 128GB+ RAM. The quantized files are available on Hugging Face with instructions for the DS4 inference engine.
@ttasanen: Just fired up DS4 by @antirez on my Mac Studio M3 Ultra 256GB and man, it’s seriously impressive. A clean, purpose-buil…
DS4 is a specialized inference engine by antirez designed to run DeepSeek V4 Flash locally on high-end Mac hardware, featuring optimized KV cache handling and 1M context support.
Follow-up: DeepSeek V4 Flash on 2x RTX PRO 6000 finishes real coding tasks faster than Sonnet and Opus, at about Sonnet quality
DeepSeek V4 Flash on dual RTX PRO 6000 GPUs completes real coding tasks faster than Anthropic's Sonnet and Opus models while achieving similar quality to Sonnet.