Search results

[Feature]: Support structured outputs for beam search

feature requestNew feature or requestNew feature or request

Status: Open.

#34782 In vllm-project/vllm;

· BrandonStudio opened on Feb 18, 2026

[Feature]: parity with cuda - ROCm Kimi K2.5 disagg PD +wideEP recipe

feature requestNew feature or requestNew feature or request

rocmRelated to AMD ROCmRelated to AMD ROCm

Status: Open.

#34781 In vllm-project/vllm;

· functionstackx opened on Feb 18, 2026

[Bug]: LlavaNext + compressed-tensors: loader expects multi_modal_projector.linear_1.weight_packed but model only has linear_1weight/linear_1bias

bugSomething isn't workingSomething isn't working

Status: Open.

#34771 In vllm-project/vllm;

· gocoding1011 opened on Feb 18, 2026

[Performance]: Remove <code>batch_size</code> and <code>max_seq_len</code> padding when cuDNN is upgraded to 9.19

performancePerformance-related issuesPerformance-related issues

Status: Open.

#34763 In vllm-project/vllm;

· maxyanghu opened on Feb 18, 2026

[Bug]: nvidia/Llama-3.3-70B-Instruct-NVFP4 Degraded / Gibberish Output with TRITON_ATTN

bugSomething isn't workingSomething isn't working

Status: Open.

#34759 In vllm-project/vllm;

· frankwang28 opened on Feb 17, 2026

Qwen3-Coder-Next-FP8 with tool calling causes system hard-freeze on multi-GPU tensor parallel (v0.15.1)

usageHow to use vllmHow to use vllm

Status: Open.

#34755 In vllm-project/vllm;

· zaidorx opened on Feb 17, 2026

[Bug]: Improve <code>--kv-cache-dtype</code> behavior when checkpoint specifies <code>kv_cache_quant_algo</code>

bugSomething isn't workingSomething isn't working

good first issueGood for newcomersGood for newcomers

Status: Open.

#34752 In vllm-project/vllm;

· pavanimajety opened on Feb 17, 2026

[Feature]: Add --force-close-connections flag to vllm bench serve

feature requestNew feature or requestNew feature or request

Status: Open.

#34746 In vllm-project/vllm;

· InfraWhisperer opened on Feb 17, 2026

[Refactor]: Handle OOV Multimodal Tokens Generically

documentationImprovements or additions to documentationImprovements or additions to documentation

Status: Open.

#34736 In vllm-project/vllm;

· alex-jw-brooks opened on Feb 17, 2026

[Bug]: Spec Decode Stats still report drafted tokens if drafting is skipped

bugSomething isn't workingSomething isn't working

Status: Open.

#34734 In vllm-project/vllm;

· benchislett opened on Feb 17, 2026

[Performance]: Improve swap_states by swapping active token prefixes

performancePerformance-related issuesPerformance-related issues

Status: Open.

#34731 In vllm-project/vllm;

· pjo256 opened on Feb 17, 2026

[Bug]: Nemotron NVFP4 Failing With TRTLLM NVFP4

bugSomething isn't workingSomething isn't working

Status: Open.

#34728 In vllm-project/vllm;

· robertgshaw2-redhat opened on Feb 17, 2026