The vllm/vllm-openai version 0.9.1 is nearly 30% faster compared to lmsysorg/sglang:v0.4.7.post, but it stops running every two to three hours

Model Support DeepSeek

start-dzl June 23, 2025, 12:51pm 1

Topic		Replies	Views	Activity
Init DeepSeek-R1 using Offline Batched Inference General	3	160	May 18, 2025
Errors When Running VLLM + DeepSeek on RTX 5090 — Existing Solutions Not Working General	1	1441	May 21, 2025
Will vLLM follow-up DeepSeek's inference system DeepSeek	3	377	May 13, 2025
The service carrying capacity and computing efficiency issues of vllm General	0	16	July 16, 2025
DeepSeek-V3 tool_choice="auto", not working but tool_choice="required" is working Tool Calling	3	139	August 25, 2025