Search results

[Bug]: “max_model_len” in “--speculative-config” is invalid

bugSomething isn't workingSomething isn't working

Status: Open.

#41456 In vllm-project/vllm;

· KevinLiuMY opened on May 1, 2026

[Bug]: Gemma4-31B-it deployed on vLLM cannot process images in tool message

bugSomething isn't workingSomething isn't working

Status: Open.

#41452 In vllm-project/vllm;

· CloudyDory opened on May 1, 2026

[Feature]: MoE Active Expert Management --moe-gpu-prefetch <num>

feature requestNew feature or requestNew feature or request

Status: Open.

#41447 In vllm-project/vllm;

· leoustc opened on May 1, 2026

[Bug]: macOS arm64 build broken — cpu_attn_vec.hpp calls FP32Vec16(const BF16Vec32&, int) constructor that doesn't exist on ARM (regression from #39445)

cpuRelated to CPU backendsRelated to CPU backends

Status: Open.

#41437 In vllm-project/vllm;

· mcsantiago opened on May 1, 2026

fused_add_rms_norm does not branch on has_weight=False (TODO(luka)); FlashNorm weightless RMSNorm cannot realize a speedup on the GPU path

Status: Open.

#41430 In vllm-project/vllm;

· fm1320 opened on May 1, 2026

Backport request: CVE-2026-24779 (urlparse → urllib3 parse_url) on releases/v0.9.0

Status: Open.

#41420 In vllm-project/vllm;

· vulgraph opened on Apr 30, 2026

[bug] spawn_new_process_for_each_test decorator broken

bugSomething isn't workingSomething isn't working

ci/build

help wantedExtra attention is neededExtra attention is needed

Status: Open.

Bug

#41415 In vllm-project/vllm;

· ProExpertProg opened on Apr 30, 2026

[Bug]: TurboQuant fails on non-power-of-2 head_dim (Phi-2, MSE-K presets)

rocmRelated to AMD ROCmRelated to AMD ROCm

Status: Open.

#41413 In vllm-project/vllm;

· TheTom opened on Apr 30, 2026

[Bug]: DSV4 TileLang JIT fails with 'Permission denied: /tmp/tvm-debug-mode-tempdirs/...' on shared multi-user hosts

bugSomething isn't workingSomething isn't working

Status: Open.

#41410 In vllm-project/vllm;

· mgoin opened on Apr 30, 2026

[Tracking] TurboQuant + Gemma 4 multimodal: 5-gate blocker stack

Status: Open.

#41403 In vllm-project/vllm;

· cferra opened on Apr 30, 2026

[Bug]: DeepSeek-V4-Flash MTP hangs with vllm bench serve when concurrency > 1 on vLLM v0.20.0

bugSomething isn't workingSomething isn't working

Status: Open.

#41402 In vllm-project/vllm;

· jasl opened on Apr 30, 2026

[Performance]: Llama-Nemotron embedding is slower than Transformers for offline batch-32 pooling after compile-cache warmup

Status: Open.

#41390 In vllm-project/vllm;

· charlesbluca opened on Apr 30, 2026