17 November 2026

Several labs report faster AI model serving and inference gains

First reported

Latent Space ran this on .

  • Together, DeepSeek and MIT researchers reported work on making AI models run faster when answering users.
  • Together describes a fused-kernel approach, a way of combining several GPU operations into one to save time.
  • DeepSeek reported throughput figures for its V4.1 Flash model on Nvidia's Vera Rubin chips.

Our read

Only one of the 30 tracked newsletters, Latent Space, covered this story, which was first reported 2026-11-17. It follows a related inference-speed story from our archive on 2026-11-16 about AMD chips.

Written by Newskeryx from which newsletters covered this story and from our own archive.

How it was covered

Latent Spaceswyx & Alessio

Kernel fusion, Vera Rubin throughput benchmarks, batch-1 latency, adaptive serving research, prefix-cache pitfalls, agent sandboxes, MLPerf training results