Baseten cuts 770GB RL weight sync to 15 seconds, 7.5x faster than prior SOTA
baseten · x · 2026-10-31
Baseten shipped what it calls the world's fastest delta weight syncs for managed RL rollouts. A 770GB GLM-5.3 checkpoint sync that previously took 36 seconds (6s request pause) now completes in 15 seconds with a 0.8-second pause — 7.5x faster than prior SOTA — so global fleets always run rollouts against the latest policy. Engineering details in their blog post.
More from Infra
- Liquid Agent brings fully local LFM3 agent to Linux with one-command install — JosephJacks_ · 2026-10-31
- a16z: GPU rental prices up 106% in 3 months while token prices fall 58% — a16z · 2026-10-31
- vLLM nightly adds PLE disk offload, with a telling [Qwen4Exp] tag on the PR — rmhubbert · 2026-10-31
- Solo AI studio: per-minute GPU rentals rendered a 328-frame Blender shot for ~$1 — r0ck3t23 · 2026-10-31
- Same 32 models benchmarked on four Macs: M5 Max matches M3 Ultra on MoE — Striking-Swim6702 · 2026-10-31
- sFNX runs 27B dense at 114-300 tok/s on 3x V100 with P2P support — jfinkpottery · 2026-10-31