Changeloglast deploy 2026-10-27
Changelog

Lanes get faster while you sleep.

Every change to the lanes, newest first. Each entry has its own link you can share.

Releases3 entries · newest first

  1. llama-3.3-70b lanes re-tuned: 412 tok/s

    Fused decode kernel on B200 lanes. +31% throughput at the same per-token price; 1,140 lanes re-tuned overnight.

  2. eu-central dedicated lanes

    Single-tenant lanes pinned to Frankfurt. Requests, weights and logs-that-don’t-exist never leave the region.

  3. billing

    Prefix caching on every lane

    90% off cached prefix tokens. On by default, nothing to change client-side.