All issues

Search results

[Bug]: /v1/messages crashes with "Unexpected item type in content" when a tool_result block contains a tool_reference item

bugSomething isn't workingSomething isn't working

Status: Open.

#52489 In vllm-project/vllm;

· zbloss opened on Aug 16, 2026

[Bug]: disagg_epd_proxy encoder fan-out restarts round-robin per request, overloading first encoder

bugSomething isn't workingSomething isn't working

Status: Open.

#52484 In vllm-project/vllm;

· frankmanbb opened on Aug 16, 2026

[Bug]: MTP speculative decoding on Qwen3.5-family models logs misleading multimodal warnings ("treated as multimodal but has no registered multimodal processor")

Status: Open.

#52481 In vllm-project/vllm;

· jianliao opened on Aug 16, 2026

[Bug]: qwen3_5_mtp fails to load at tensor-parallel-size >= 2 (drafter weight shape mismatch)

quantization

Status: Open.

#52480 In vllm-project/vllm;

· sahaq opened on Aug 16, 2026

[Bug]: Sleep-mode Level 2 wake does not reload speculative-decoding draft weights — silent ~2x decode slowdown, acceptance drops to 0

quantization

Status: Open.

#52479 In vllm-project/vllm;

· Suppressor72 opened on Aug 16, 2026

[Bug]: MTP speculative decoding produces repetition collapse with turboquant_* KV cache on sm120 (Qwen3.8-27B hybrid GDN)

quantization

Status: Open.

#52475 In vllm-project/vllm;

· mechramc opened on Aug 15, 2026

Streaming input: per-chunk arrival cost is O(cumulative prompt), making long multimodal sessions quadratic

Status: Open.

#52471 In vllm-project/vllm;

· sseanliu opened on Aug 15, 2026

[Bug][DP] Tie-break rotation from #47420 destroys prefix-cache locality for multi-turn traffic under internal DP LB

Status: Open.

#52463 In vllm-project/vllm;

· kfastino opened on Aug 15, 2026

[Bug]: Under async scheduling, stateful logits processors that read output_token_ids values see -1 placeholders — logitsprocs_need_output_token_ids only counts CLI custom processors

Status: Open.

#52461 In vllm-project/vllm;

· danbo610 opened on Aug 15, 2026

[Bug]: 0.27.1 crashes in custom_all_reduce (illegal memory access) during CUDA graph capture on Hopper TP=4; identical config works on 0.26.0

Status: Open.

#52457 In vllm-project/vllm;

· Robinwm049 opened on Aug 15, 2026

[Bug]: Qwen 3.8 quark int4 loading failed.

bugSomething isn't workingSomething isn't working

quantization

rocmRelated to AMD ROCmRelated to AMD ROCm

Status: Open.

#52454 In vllm-project/vllm;

· hnhyzz opened on Aug 15, 2026

[Bug]: the issues about install vllm with cuda 12.6, Why hasn't anyone answered this question?

bugSomething isn't workingSomething isn't working

Status: Open.

#52453 In vllm-project/vllm;

· cqray1990 opened on Aug 15, 2026