Issues

• Assess native GenAI execution of split Qwen4-Exp independently from export

  onnxruntime/mobius#775 · Filter by author titaiwangms opened on Oct 14, 2026

• Execute staged official-weight Qwen4-Exp qualification under explicit budgets

  onnxruntime/mobius#774 · Filter by author titaiwangms opened on Oct 14, 2026

• Implement reduced greedy speculation with complete Qwen4-Exp state transactions

  onnxruntime/mobius#773 · Filter by author titaiwangms opened on Oct 14, 2026

• Implement independently referenced standard-ONNX Qwen4-Exp MTP forward

  onnxruntime/mobius#772 · Filter by author titaiwangms opened on Oct 14, 2026

• Qualify reduced Qwen4-Exp split generation with real token feedback

  onnxruntime/mobius#771 · Filter by author titaiwangms opened on Oct 14, 2026

• Implement independently loadable standard-ONNX Qwen4-Exp ngram/main packages

  onnxruntime/mobius#770 · Filter by author titaiwangms opened on Oct 14, 2026

• Ground Qwen4-Exp MTP equations, token alignment, cache ABI, and reference tests

  onnxruntime/mobius#769 · Filter by author titaiwangms opened on Oct 14, 2026
  

• Design the opt-in Qwen4-Exp ngram/main API, state ABI, and weight contract

  onnxruntime/mobius#768 · Filter by author titaiwangms opened on Oct 14, 2026
  

• Qwen3.8-Flash-Next: standard-ONNX ngram/main separation, MTP, and staged qualification

  onnxruntime/mobius#767 · Filter by author titaiwangms opened on Oct 13, 2026
  

• Qwen3VL export has inconsistent dtype for visual_encoder

  onnxruntime/mobius#750 · Filter by author vigneshm2 opened on Sep 30, 2026

• Add a CPU runtime contract test for mixed-width QMoE

  onnxruntime/mobius#745 · Filter by author titaiwangms opened on Sep 22, 2026
  

• Add Qwen3.8 packed hybrid PagedAttention export for ORT GenAI Engine

  aienhancementtracking
  onnxruntime/mobius#737 · Filter by author titaiwangms opened on Sep 16, 2026
  

• Support per-layer Olive mixed-precision QMoE export

  onnxruntime/mobius#735 · Filter by author titaiwangms opened on Sep 14, 2026
  

• Track official VibeVoice 1.5B HF sidecars

  tracking
  onnxruntime/mobius#727 · Filter by author justinchuby opened on Sep 4, 2026

• [Model Request]: Enable Qwen4Exp GGUF payload import and execution

  model-request
  onnxruntime/mobius#699 · Filter by author justinchuby opened on Aug 29, 2026

• Clarify GGUF quantized storage fidelity and BitShift fallback compute

  onnxruntime/mobius#667 · Filter by author justinchuby opened on Aug 27, 2026

• Close remaining heterogeneous recurrent-state runtime gaps

  onnxruntime/mobius#605 · Filter by author justinchuby opened on Aug 25, 2026

• [Model Request]: nvidia/magpie_tts_multilingual_357m

  model-request
  onnxruntime/mobius#570 · Filter by author justinchuby opened on Aug 23, 2026

• 4bit diffusion models

  onnxruntime/mobius#569 · Filter by author justinchuby opened on Aug 23, 2026

• Weekly prompt

  onnxruntime/mobius#481 · Filter by author justinchuby opened on Aug 14, 2026

• Convert weather next

  onnxruntime/mobius#468 · Filter by author justinchuby opened on Aug 10, 2026
  

• Plans to Support VibeVoice/Nemotron - TTS Support

  onnxruntime/mobius#356 · Filter by author leestott opened on Jun 15, 2026
  2

  

• Emit opset-24 external KV cache (nonpad_kv_seqlen) for bias decoders to recover near-GQA decode (Gemma-4, sliding-window)

  onnxruntime/mobius#349 · Filter by author justinchuby opened on Jun 9, 2026
  1

  

• Benchmark: Gemma4 ORT GenAI inference — f16 vs Q4_K_M vs NF4 (CUDA + CPU)

  onnxruntime/mobius#346 · Filter by author justinchuby opened on Jun 8, 2026
  1

• Emit maskless is_causal=1 + nonpad_kv_seqlen static-cache graph once upstream lands

  onnxruntime/mobius#345 · Filter by author titaiwangms opened on Jun 8, 2026