NNsight 0.8 pre-release ships faster engine, MoE and near-native vLLM support

davidbau · x · 2026-09-10

NDIF's NNsight 0.8 is out in pre-release for interpretability researchers. The new version keeps the same API while shipping a rewritten execution engine with better support for MoE transformers and near-native-throughput vLLM interpretability.

The team welcomes feedback and feature requests during pre-release.

Original post →

More from Research

Research channel →