Search results

[Bug]: Qwen2-Audio inference broken on transformers >= 5.x — generates only <|audio_bos|> / <|audio_eos|> instead of text

Status: Open.

#46250 In vllm-project/vllm;

· wentian-byte opened on Jun 20, 2026

[Bug]: [Regression] Qwen3.6-27B tool calls fail on Responses API when MTP is enabled

bugSomething isn't workingSomething isn't working

Status: Open.

#46249 In vllm-project/vllm;

· rmhubbert opened on Jun 20, 2026

v1 P/D: _update_from_kv_xfer_finished AssertionError (kills EngineCore) when an aborted request's finished_recving + finished_sending land in the same step

Status: Open.

#46240 In vllm-project/vllm;

· S1ro1 opened on Jun 20, 2026

[Bug]: W4A16 or W8A16 Qwen3.5 9B meet AssertionError

bugSomething isn't workingSomething isn't working

Status: Open.

#46230 In vllm-project/vllm;

· whyseu opened on Jun 20, 2026

[Bug]: Streaming parser engine leaks special/structural tokens (BOS/EOS/drop_tokens) in non-streaming responses

Status: Open.

#46224 In vllm-project/vllm;

· fank opened on Jun 20, 2026

[Feature]: Add a 2-bit KV-cache quantisation backend from Together AI

feature requestNew feature or requestNew feature or request

Status: Open.

#46221 In vllm-project/vllm;

· zhangj1an opened on Jun 20, 2026

[Bug]: MiniMax‑M3 (MSA) + NixlConnector: connector forces HND KV layout while MSA is NHD‑native, permuting the head axis under P/D disaggregation

bugSomething isn't workingSomething isn't working

Status: Open.

#46204 In vllm-project/vllm;

· biswapanda opened on Jun 20, 2026

[Bug]: OTEL "Dynamo bytecode transform" span reports wildly incorrect duration (tens of thousands of days)

Status: Open.

#46193 In vllm-project/vllm;

· nicklasfrahm opened on Jun 19, 2026

[Feature]: Faster hybrid-SSM decode via input "replay" ring buffer (ReplaySSM)

feature requestNew feature or requestNew feature or request

speculative-decoding

Status: Open.

#46187 In vllm-project/vllm;

· mgoin opened on Jun 19, 2026

RFC: State-Aware Monolingual Reasoning Constraints for DeepSeek-R1 Models (Opt-In)

Status: Open.

#46185 In vllm-project/vllm;

· Himanshu-is-code opened on Jun 19, 2026

[Bug]: vllm serve neuralmagic/Llama-3.2-1B-quantized.w8a8 Failing On ROCm w/ MRV2

bugSomething isn't workingSomething isn't working

rocmRelated to AMD ROCmRelated to AMD ROCm

Status: Open.

#46179 In vllm-project/vllm;

· micah-wil opened on Jun 19, 2026

[Bug]: Segfault in DNNL matmul during mixed-batch GQA after #43032 (v0.23.0 regression)

bugSomething isn't workingSomething isn't working

Status: Open.

#46131 In vllm-project/vllm;

· vinay-murakala opened on Jun 19, 2026