Skip to main content
← Back to news

Tideway: VRAM and RAM managed as one pool for local MoE inference

Research
Paper on Hugging Face: TIDEWAY treats system RAM and GPU VRAM as one memory pool, with layouts adapted to the prefill and decode phases. On consumer hardware it claims 1.97x end-to-end speedup and 2.40x on time-to-first-token over the fastest baseline.

Get the next local-AI briefing

Once a week, the top selection from iaLocale: articles, news and resources on local AI, and more.

After signing up you receive a confirmation email (double opt-in). Without confirmation we send nothing. No spam, we promise.