v0.30.0 v0.28.0
RyanCodrai/turbovecSep 25, 2026 → Aug 31, 20260 days between releases
Generating AI comparison
Analyzing what changed between v0.30.0 and v0.28.0. This page updates automatically — usually within a minute.
The Two Releases
v0.30.0
Sep 25, 2026· 26 days ago
This release features 504 merged changes and introduces a unified full-duplex serving framework, native cross-stage KV transfer, and interactive world-model video generation. It supports realtime Minimax H3 inference on Nvidia Blackwell and Ascend NPU 950, rebasing onto vLLM 0.30.0 with expanded model and hardware coverage.
Full release notesv0.28.0
Aug 31, 2026· 2 months ago
This release features 397 merged changes, rebasing onto vLLM 0.28. It introduces scheduler-managed paged KV cache for diffusion, production-level serving for Minimax-H3, and expands the full-duplex speech stack with PersonaPlex and Nemotron VoiceChat.
Full release notes