v2.0.8-beta v0.30.0
MiniMax-AI/MiniMax-01Sep 21, 2026 → Sep 22, 20260 days between releases
What's New
Major update introduces Fast Start caching, Model Runner V2, and new models with breaking config changes.
Breaking Changes
- Scale-out endpoints now require the `--enable-scale-out` flag
- GPTQ activation ordering (`g_idx`) has been removed
- Deprecated environment variables have been removed
- Deprecated `python -m vllm.entrypoints.grpc_server` command has been removed
New Features
- DeepSeek-V4.1-Flash and other new model support
- Fast Start weight caching with persistent daemon
- Gumbel-max watermarking generation and detection
- HiSparse host-resident tier for sparse-MLA
- Model Runner V2 with dual-batch overlap and CUDA graphs
Upgrade Notes
Scale-out endpoints now require the `--enable-scale-out` flag. The GPTQ activation ordering `g_idx` and deprecated environment variables are no longer supported.
The Two Releases
v2.0.8-beta
Sep 21, 2026· 1 month ago
This release focuses on critical security updates regarding JWT authentication and OnlyOffice callbacks, alongside UI responsiveness improvements and bug fixes for public shares and downloads.
Full release notesv0.30.0
Sep 22, 2026· 30 days ago
A major performance and feature release introducing new models, Fast Start caching, watermarking, HiSparse, and Model Runner V2, along with significant breaking changes to configuration.
Full release notes