Repository Issue Activity (beta)

#37308 [Bug]: Severe Head-of-Line Blocking (147x TTFT) under Prefix Caching with Asymmetric Batches

#40076 [Feature]: Per-request timing metrics in response body

#47648 [Bug]: DeepSeek-V4-Flash-DSpark fails on H200/SM90 with FlashMLA KV cache shape mismatch

#47159 [Bug]: triton_w4a16_gemm assert qzeros.shape == (K // group_size, N // 8) fails on GPTQ models

#41663 [Bug]: XPU TP=2 on dual Intel Arc Pro B70 (Battlemage): GP fault + xe BCS engine reset reproduces in intel/vllm:0.17.0-xpu on Ubuntu 24.04 HWE 6.17

#40554 [AMD][CI Failure][Tracker] Static dashboard tracker for current CI failures

#46526 [RFC]: Add Helion linear backend for vLLM

#46856 [Bug][ROCm] GLM5.1 ROCM_AITER_MLA_SPARSE corruption persists between ISL 60k to 137k (AA-LCR dataset) on gfx942 in v0.23.0 (both compiled and eager mode)

#47768 [RFC]: vLLM gRPC server API for downstream integrations

#47769 [Bug]: MXFP4 MoE Marlin fallback crashes with illegal memory access on Hopper (DeepSeek V4, TP+EP) — non-tile-aligned intermediate size

#46107 [RFC]: Heterogeneous TP<->DP parallelism across the P/D disaggregation boundary (MoRIIO)

#45608 [New Model]: gemma4-12b-qat-w4a16

#47763 [Performance]: MiniMax-M3 verify-as-decode re-streams the sparse KV from HBM once per draft token

#36180 [Bug]: meta-llama/Llama-3.2-1B-Instruct Fails With ROCM_ATTN Due To Seg Fault

#47172 Model Runner V2 Remaining TODOs

#47456 [RFC]: Design a shared warmup infrastructure for JITs

#43475 [Feature]: Add SpectralQuant KV Cache Compression (builds on TurboQuant #38479)

#38930 Entropy-adaptive per-head KV cache quantization: +8% quality over uniform at same compression

#46985 [Bug]: Why does kv-cache-dtype=fp8 OOM more easily than bf16/fp16 on long-context GLM-5.1-AWQ runs?

#46221 [Feature]: Add a 2-bit KV-cache quantisation backend from Together AI

#46710 [Bug] DeepSeekV4-Flash produces incorrect output with inline system messages after PR #46025 when `preserved in-place`

#47761 [Bug]: vllm 0.23.0 and 0.24.0 - Qwen3.6-35B-A3B-FP8 - Fails generating code- "400 Unterminated string starting at"

#38171 [Feature]: Add TurboQuant Support for KV Cache Quantization

#46078 Preserve FP8 indexer WK pairs across incremental load_weights calls

#41283 [Bug] Gemma 4 31B crashes on k_eq_v full-attention layers (QKV split shape mismatch)