llama-modchip runs 284B DeepSeek on 64GB RAM, no GPU

Open-source project llama-modchip released v2.4.0, claiming 284B DeepSeek-V4-Flash can run entirely in RAM on a 64GB desktop or server with zero GPUs, using a 56.56 GiB checkpoint. It reports 44.36 tok/s prefill and 9.1-14.7 tok/s generation at 256K context, pitching CPU-only MoE inference for local workstations.

• Kimi K3.1 Pro Open Weights Released2026-11-04
• DeepSeek releases enhanced coding agent to rival Anthropic2026-11-01
• DeepSeek Ships Minor V4.1-Flash-1017 Update2026-10-17
• South Korean banks hacked, possibly by one person using AI tools2026-10-08
• Claude Code 'Don't Ask Again' Exposes Passwords in 1,122 Public GitHub Files2026-11-07