AION
Open-source releaseEfficiency & Inference · MLOps, Tooling & Infrastructure1 source · May 15, 2026

vllm-project/vllm v0.21.0

Transformers v4 deprecated: This release formally deprecates transformers v4 support (#40389).

NVIDIADeepSeekQwenDeepSeek modelsPyTorchvLLM
← Back to brief

Key points

  • This release features 367 commits from 202 contributors (49 new)!
  • C++20 build requirement: vLLM now requires a C++20-compatible compiler for compatibility with PyTorch (#40380).
  • New architectures: MiMo-V2.5 (#40967), Laguna XS.2 (#41129, #41880), Moondream3 (#32325), Qianfan-OCR (#40136), Cohere MoE (#40817), Cohere Eagle (#42078).
  • DeepSeek V4: AMD/ROCm support (#40871), pipeline parallelism (#41694), max reasoning effort (#40982), disaggregated serving fixes (#41957).

Sources (1)

  • [1]vllm-project/vllm v0.21.0
    GitHub: vllm-project/vllm · May 15, 08:44 AM
    * **Transformers v4 deprecated**: This release formally deprecates `transformers` v4 support (#40389).
    This release features 367 commits from 202 contributors (49 new)!

Extractive summary: sentences quoted from the sources.