v0.19.1

vllm-project/vllmv0.19.1Apr 18, 2026by khluu

AI Summary

Patch release on top of v0.19.0 with Transformers v5.5.3 upgrade and comprehensive bug fixes for Gemma4 including streaming tool calls, HTML duplication, and token repetition issues.

Key Highlights

Update to transformers v5.5.3
Gemma4 streaming tool call fixes (invalid JSON, HTML duplication, split boolean/number values)
Gemma4 quantized MoE support
Gemma4 Eagle3 speculative decoding support
Fix Gemma4 token repetition via dynamic BOS injection

New Features

Gemma4ForCasualLM LoRA adapter loading fix
Gemma4 tool parser null value handling fix
Reasoning parser integration with adjust_request tool

Full Release Notes

This is a patch release on top of `v0.19.0` with Transformers v5.5.3 upgrade and bug fixes for Gemma4:
- Update to transformers v5 (#30566)
- [Bugfix] Fix invalid JSON in Gemma 4 streaming tool calls by stripping partial delimiters (#38992)
- [Bugfix][Frontend] Fix Gemma4 streaming HTML duplication after tool calls (#38909)
- [Bugfix] Fix Gemma4 streaming tool call corruption for split boolean/number values (#39114)
- [Tool] adjust_request to reasoning parser, and Gemma4 fixes (#39027)
- [Gemma4] Support quantized MoE (#39045)
- Add Gemma4 Eagle3 support (#39450)
- [Gemma4][Bugfix]: Enable Gemma4ForCasualLM to load lora adapters correctly (#38844)
- [Bugfix] Fix Gemma4 tool parser converting bare null to string "null" (#39679)
- [Model] Fix Gemma 4 token repetition by dynamic BOS injection for PT models (#39842)
- fix(kimi_k25): resolve media_placeholder_token_id from tokenizer (#39344)