Source: vLLM release notes v0.8.0 and v0.11.0

GitHub release notes marking V1 as the default engine (v0.8.0, March 2025) and the complete removal of V0 (v0.11.0, October 2025).

Contents
  • What: GitHub release notes (two releases)
  • Author: vLLM project
  • Date: v0.8.0 2025-03-18; v0.11.0 2025-10-02
  • Original: v0.8.0 release · v0.11.0 release

Summary

Two milestones in the V0-to-V1 transition. v0.8.0 switched the default engine to V1 for supported use cases, with an environment-variable escape hatch. v0.11.0 deleted the V0 engine code, leaving V1 as the only engine.

Key claims

  • v0.8.0: V1 enabled by default for supported cases; VLLM_USE_V1=0 opts out, and users doing so were asked to file an issue.
  • v0.11.0: completes V0 removal (V0 engine classes and attention backends gone); V1 is the only engine in the codebase; CUDA graph mode default becomes FULL_AND_PIECEWISE.

Feeds

vLLM

Suggest changes →