DeepSeek V4 MLX 8-bit Gets Local Inference Speedup
No_Run8812 · reddit · 2026-07-06
A user optimized the DeepSeek V4 Flash 8-bit MLX model on oMLX using Codex, achieving around 1.6x prefill and 3x decode speedups. They utilized the 302GB 8-bit affine MLX version from mlx-community to avoid the precision loss associated with 4-bit. The author, noting they are not a Metal kernel expert, shared the modification details for community review, confirming that tool-calling currently works normally with reportedly no loss in precision.
More from Infra
- Analyst: Ant has the most fungible AI workload of any model provider today — BenBajarin · 2026-10-26
- Anthropic S-1: training infra edge cut per-task cost 40% and doubled revenue growth — BenBajarin · 2026-10-26
- Rewriting scaling law study: optimal synthetic data share rises from 10% to 32% with compute — heghbalz · 2026-10-26
- US states halt new data centers as 72% of Americans oppose local builds — YvesMulkers · 2026-10-26
- $5M Seed Round Can't Afford One DeepSeek Training Run: GPU-less AI Startup's Three Paths — FinanceYF5 · 2026-10-26
- HBM demand shifts: low-stack inference plans shelved, Rubin to go 12-high only — BenBajarin · 2026-10-26