Issues

• [Bug] /v1/realtime history transcription inherits the reply's token limit

  sgl-project/sglang-omni#2463 · Filter by author twu3202 opened on Oct 1, 2026

• [Bug] MiniCPM-o /v1/audio/transcriptions answers spoken instructions instead of transcribing them

  sgl-project/sglang-omni#2462 · Filter by author twu3202 opened on Oct 1, 2026

• [Roadmap] Optimization for Nemotron-VoiceChat 11b

  sgl-project/sglang-omni#2459 · Filter by author lijrjyan opened on Oct 1, 2026
  9

• [Roadmap] Optimization for PersonaPlex

  sgl-project/sglang-omni#2458 · Filter by author lijrjyan opened on Oct 1, 2026
  5

• [performance optimization] reduce image pre-processing cost for mini cpm

  sgl-project/sglang-omni#2453 · Filter by author JINO-ROHIT opened on Sep 30, 2026
  

• [Feature] Qwen3-TTS: support configurable output sample rate in the Speech API

  sgl-project/sglang-omni#2448 · Filter by author Shuzhenzhou123 opened on Sep 30, 2026

• [Roadmap][AMD] AMD ROCm Support Roadmap

  sgl-project/sglang-omni#2416 · Filter by author zijiecode opened on Sep 28, 2026

• [Runtime Profiling] MiniCPM-o 4.5 speech pipeline

  sgl-project/sglang-omni#2399 · Filter by author Dayuxiaoshui opened on Sep 27, 2026

• [Bug] [CPU] Importing sglang_omni.scheduling.sglang_backend SIGILL on non-AVX-512-BF16 hosts

  sgl-project/sglang-omni#2384 · Filter by author notnotharsh opened on Sep 26, 2026

• [Tracking][Qwen3-Omni] Serving Reform and Default Configuration Roadmap

  sgl-project/sglang-omni#2362 · Filter by author AkazaAkane opened on Sep 24, 2026
  

• [Feature] Qwen3-TTS: accept a precomputed speaker embedding instead of ref_audio for Base cloning

  sgl-project/sglang-omni#2310 · Filter by author farzanehnakhaee70 opened on Sep 22, 2026
  1

• [Docs] Local preview commands depend on the current working directory

  sgl-project/sglang-omni#2301 · Filter by author Tao20060602 opened on Sep 22, 2026
  2

• [Qwen3-TTS] Streaming first audio, current round: vocoder placement, Base first-audio path, latency CI

  sgl-project/sglang-omni#2294 · Filter by author luojiaxuan opened on Sep 22, 2026

• [Tracking] MiniCPM-o4.5 end-to-end inference optimization

  sgl-project/sglang-omni#2284 · Filter by author MayDomine opened on Sep 21, 2026
  7

• [Runtime Profiling] MiniCPMo-4.5

  sgl-project/sglang-omni#2273 · Filter by author BruceLoveDecimal opened on Sep 20, 2026
  8

• [Eval] add a replayable native full-duplex benchmark

  enhancement
  sgl-project/sglang-omni#2270 · Filter by author edwingao28 opened on Sep 20, 2026
  1

• [Qwen3-Omni] Audio encoder startup fails after Transformers helper rename

  sgl-project/sglang-omni#2269 · Filter by author edwingao28 opened on Sep 20, 2026

• [CI] Manual nightly run reports

  sgl-project/sglang-omni#2246 · Filter by author edwingao28 opened on Sep 18, 2026

• [Feature] In-app HF_ENDPOINT for first-time ASR download

  sgl-project/sglang-omni#2239 · Filter by author fengxiaohu opened on Sep 18, 2026
  1

• [Feature Request] Support release_memory_occupation / resume_memory_occupation for Omni stages

  sgl-project/sglang-omni#2236 · Filter by author Hana19951208 opened on Sep 18, 2026
  6

• [RFC] Core AI POC vs MLX-mini on Apple Silicon

  sgl-project/sglang-omni#2223 · Filter by author fengxiaohu opened on Sep 17, 2026
  2

• [Feature] Enable torch.compile by default with persistent first-start cache

  enhancement
  sgl-project/sglang-omni#2211 · Filter by author zhaochenyang20 opened on Sep 16, 2026

• [Feature] Bound conversation history growth in /v1/realtime sessions

  sgl-project/sglang-omni#2208 · Filter by author izhl opened on Sep 16, 2026
  2

• [Roadmap] LLaDA-Uni Model Support, Inference Optimization and Serving

  sgl-project/sglang-omni#2207 · Filter by author Anmuliar opened on Sep 16, 2026
  4

• [Roadmap] Coding Style

  sgl-project/sglang-omni#2195 · Filter by author SunskyXH opened on Sep 15, 2026
  5