vllm

[Model] Add DeepSeek-V3.1 reasoning parser (split from PR #24972)

#25589

Merged