Search results

[Bug]: Performance regression from v0.16.0 to v0.17.0+ on openai/gpt-oss-120b

bugSomething isn't workingSomething isn't working

Status: Open.

#40838 In vllm-project/vllm;

· aas008 opened on Apr 24, 2026

[Bug]: TurboQuant KV × any speculative decoding (MTP or ngram) produces degenerate token loops — confirmed across dense and hybrid attention

Status: Open.

#40831 In vllm-project/vllm;

· noonghunna opened on Apr 24, 2026

[RFC]: [RL - WIP] Checkpoint -> Kernel Format transition graph

RFC

Status: Open.

#40822 In vllm-project/vllm;

· S1ro1 opened on Apr 24, 2026

[Bug]: Deepseek V4 failed to load on RTX PRO 6000

bugSomething isn't workingSomething isn't working

Status: Open.

#40821 In vllm-project/vllm;

· jerin-metrum-ai opened on Apr 24, 2026

[Bug]: Qwen3.6 streaming chat completions emit final answer in delta.reasoning and leave delta.content empty even with enable_thinking=false

bugSomething isn't workingSomething isn't working

Status: Open.

#40816 In vllm-project/vllm;

· xy3xy3 opened on Apr 24, 2026

[Feature]: Support a node-local external DP mode with a single vllm serve and aggregated admin health endpoint

feature requestNew feature or requestNew feature or request

Status: Open.

#40814 In vllm-project/vllm;

· yewentao256 opened on Apr 24, 2026

[Feature]: Add a standardized config extension point for out-of-tree platforms

feature requestNew feature or requestNew feature or request

Status: Open.

#40813 In vllm-project/vllm;

· MTesla-X opened on Apr 24, 2026

[Feature]: Support DDTree speculative decoding

feature requestNew feature or requestNew feature or request

Status: Open.

#40809 In vllm-project/vllm;

· ivanbaldo opened on Apr 24, 2026

[Bug]: TurboQuant KV + spec-decode + chunked-prefill crashes CUDA graph capture at query_start_loc.tolist() in continuation-prefill path (Qwen3-Next hybrid dense)

bugSomething isn't workingSomething isn't working

Status: Open.

#40807 In vllm-project/vllm;

· noonghunna opened on Apr 24, 2026

[Feature]: Deepseek V4 cannot run ,Please support SM120 GPU,example rtx5090 rtxpro6000

feature requestNew feature or requestNew feature or request

Status: Open.

#40802 In vllm-project/vllm;

· wuwenthink opened on Apr 24, 2026

[Bug]: Title: DeepSeek V4 intermittently leaks DSML fragments in auto + streaming mode, causing unstable tool-call parsing

bugSomething isn't workingSomething isn't working

Status: Open.

#40801 In vllm-project/vllm;

· Windswithyou opened on Apr 24, 2026

[Bug]: Workspace allocation failure when combining Decode Context Parallelism (DCP) with EAGLE3 speculative decoding

bugSomething isn't workingSomething isn't working

Status: Open.

#40791 In vllm-project/vllm;

· BenWongCityuCS opened on Apr 24, 2026