llama-modchip Runs 284B DeepSeek on 64GB RAM With No GPU

1 reports archived1 independent sourcesupdated 11/8/2026, 18:22:04

Synthesis & Latest Updates

1 Sources Cross-Validated

The open-source llama-modchip v2.4.0 claims it can run a 284B-parameter DeepSeek-V4-Flash MoE model entirely in RAM on a 64GB CPU-only workstation, peaking at 57.61 GiB with roughly 9-15 tok/s decode at 256K context. It relies on quantization and memory optimizations such as SwiGLU-H2, plus Fractal-64/MXFP4 transcoding and status tooling.

LATEST/The open-source llama-modchip v2.4.0 claims it can run a 284B-parameter DeepSeek-V4-Flash MoE model entirely in RAM on a 64GB CPU-only workstation, peaking at 57.61 GiB with roughly 9-15 tok/s decode at 256K context. It relies on quantization and memory optimizations such as SwiGLU-H2, plus Fractal-64/MXFP4 transcoding and status tooling.

Hourly heat curve

3 fully observed hours

Now

13.2

Peak

13.619:00

24h change

–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 1 reports · Latest first

Hacker News AIT2Nov 8, 18:22·74 pts
llama-modchip Runs 284B DeepSeek on 64GB RAM With No GPU
- 284B DeepSeek-V4-Flash runs CPU-only on 64GB RAM, peaking near 57.61 GiB
- Decode is about 9.1-14.7 tok/s at 256K context, with prefill near 44 tok/s
- Ships Fractal-64 and Fractal-MXFP4 quantization transcoding plus CPU readiness checks
View Related Reports & Analysis Original Source