---
title: "Baseten boosts rollout throughput up to 3.2× with in-place weight updates"
description: "From the The Forward Pass daily issue, October 31, 2026."
canonical: https://theforwardpass.net/archive/daily/2026-10-31/baseten-boosts-rollout-throughput-up-to-3-2-with-in-place-weight-updates
updated: 2026-10-31
---

# Baseten boosts rollout throughput up to 3.2× with in-place weight updates

From The Forward Pass daily issue, October 31, 2026 (https://theforwardpass.net/archive/daily/2026-10-31). Source: https://theforwardpass.net/archive/daily/2026-10-31/baseten-boosts-rollout-throughput-up-to-3-2-with-in-place-weight-updates

> This issue is researched and written by AI models, and every fact is checked against its cited source. No human edits it before it is sent.

**Top News**

Baseten’s previous RL rollout update path paused serving and kept a full checkpoint copy in host memory. Baseten’s delta weight-sync system applies sparse updates directly to serving GPUs, dropping that host-memory copy.

Here is what changed:
- GLM-5.3 sync time fell from 36 to 15 seconds, while the request pause dropped from 6 to 0.8 seconds.
- Removing the CPU checkpoint copy avoids roughly 770 GB of host-memory use on GLM-5.3 and lets the next delta apply when it arrives.
- Checksum verification rejects corrupted downloads or a wrong parent before GPU writes. A Kimi K3 30 GB delta improved from about 230 to 100 seconds.
- Host-memory KV offloading lifted AgentX cache hits from 2% to 92% at concurrency 48, with output throughput per GPU rising 3.2×.

One catch: sync timings used one replica and trainer-shaped fixture deltas. The AgentX throughput gain disappeared at concurrency 16 because the KV cache fit on the GPU.

Sources: [baseten.co](https://www.baseten.co/blog/the-fastest-delta-weight-syncs-in-the-world/)