# Hugging Face — Shipping a Trillion Parameters With a Hub Bucket: Delta Weight Sync in TRL

- Company: Hugging Face (huggingface.co)
- Announced: 2026-05-27T00:00:00+00:00
- Category: developer-tool-release
- Subject: Platform
- Models affected: Qwen3-0.6B
- Source: https://huggingface.co/blog/delta-weight-sync
- Record: https://forck.live/items/1498-shipping-a-trillion-parameters-with-a-hub-bucket-delta-weight-sync-in-trl

Hugging Face announces a new delta weight sync feature in TRL that encodes only the changed elements of model weights between RL optimizer steps as a sparse safetensors file, uploads it to a Hugging Face Bucket, and tells vLLM to fetch it, reducing per-step payload from 1.2 GB to 20-35 MB for Qwen3-0.6B.

## Evidence

Verbatim from https://huggingface.co/blog/delta-weight-sync:

> We landed a TRL PR that encodes just the changed elements as a sparse safetensors file, uploads it to a Hugging Face Bucket, and tells vLLM to fetch it. On Qwen3-0.6B, the per-step payload drops from 1.2 GB to 20 to 35 MB.

---

Record: https://forck.live/items/1498-shipping-a-trillion-parameters-with-a-hub-bucket-delta-weight-sync-in-trl
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md