Research2026-09-03

Prime Intellect rebuilt how its training stack copies model weights to inference servers. The new path uses NIXL, a direct memory access library from Nvidia's Dynamo team, plus its own ModelExpress. Copying a 1.6 TB GLM-5.2 policy now takes under four seconds instead of 86 seconds. The system traces how vLLM lays out weights at runtime, so no mapping is written by hand. Transfer of one group of weights overlaps compression of the previous group. Most of the remaining delay is vLLM waiting for all workers to agree to pause. Prime Intellect changed that agreement to happen every wave rather than every 32. It is working with the vLLM team to check whether this hurts serving throughput. The older method required a fixed set of processes, which blocked fault tolerance and auto-scaling. Auto-scaling on the new foundation is not yet built.

What changed

Weight sync from trainer to inference took 60-90 seconds using NCCL broadcasts.

What it unlocks

Syncing a trillion-parameter policy to inference workers in under four seconds, without a fixed set of processes.

  • 86.1s to 3.9s per weight sync
  • 1.6 TB policy transferred
  • 45 GB/s of 50 GB/s NIC peak

What you need to act on it

  • prime-rl and vLLM
  • RDMA-capable GPU cluster with per-GPU 400Gb/s NICs
  • per-wave pause consensus in vLLM

Send this to someone who needs it

Shares the story and its sources. Nothing about you.

What does this mean for your job?

This is the story as everyone gets it. Once a week we send you the version written for your role — what changed, why it matters for the work you actually do, and one thing to try. Free while we tune it.