Liquid Inference routing hits >90% average cache hit-rate, best on DeepSeek and GLM
AccBalanced · x · 2026-11-13
Brett Harrison reports that the Liquid Inference routing algorithm has been optimized for cache hit-rate, now averaging over 90% across models, with even higher rates on DeepSeek and GLM providers. A companion screenshot shows DeepSeek V4.1 Flash providers at 97-100% cache hit and GLM 5.3 Flash at 87-96%, which translates to cheaper cached-token pricing and faster responses for agent-heavy workloads.
More from Infra
- Cerebras Q3: cloud revenue up 272%, 700MW data center capacity under contract — Sethwinterroth · 2026-11-13
- Critic faults chip export-control research: ByteDance lines up 36,000 B200s in Malaysia — nitarshan · 2026-11-13
- Tracked GPU acquisition likened to SR-71's Soviet titanium, sourced via hundreds of willing neoclouds — bookwormengr · 2026-11-13
- DynaDFlash: diffusion drafter boosts speculative decoding acceptance length by 5% — charles_irl · 2026-11-13
- Report: Nvidia planning frontier-class open models, sources claim — StewartalsopIII · 2026-11-13
- Celesto CI runs GitHub Actions on microVMs for ~90% less per minute — aniketmaurya · 2026-11-13