We cut model cold-start times from minutes to seconds. 60x faster, saving thousands of GPU-minutes and hundreds of terabytes of transfers ev…

We cut model cold-start times from minutes to seconds. 60x faster, saving thousands of GPU-minutes and hundreds of terabytes of transfers every day. It turns out GPUs already holding weights are faster weight servers than cloud storage. The writeup covers how we built around that, including the failure modes we hit along the way. https://runwayml.com/news/60x-faster-cold-starts-treating-peer-gpus-as-weight-servers

Source: Cristobal Valenzuela (X) | 2026-05-05