Issues

• [Performance] [ONNX Runtime] TransposeOptimizer misses inverse transposes when perm is omitted

  performance
  microsoft/onnxruntime#34008 · Filter by author nikokurumi opened on Nov 9, 2026

• [Performance] [ONNX Runtime] LayerNormFusion misses rank-one normalization with axis 0

  performance
  microsoft/onnxruntime#34006 · Filter by author miyorena opened on Nov 9, 2026

• [CPU] ReduceMax, ReduceMin, ArgMax and ArgMin ignore NaN unless it is the first element

  microsoft/onnxruntime#34001 · Filter by author 14H034160212 opened on Nov 9, 2026

• CPU: FP16 MatMul followed by Cast to FP32 bypasses FP16 rounding

  microsoft/onnxruntime#33998 · Filter by author tangyixuan01 opened on Nov 9, 2026
  1

• [Performance] CUDA GroupQueryAttention: each new prompt length waits for a cuDNN graph build (default sdpa_kernel, SM 12.0)

  ep:CUDA
  microsoft/onnxruntime#33995 · Filter by author john-rocky opened on Nov 9, 2026
  1

• [MIGraphX EP] Model cache reuses a compiled program for a different input shape

  ep:MIGraphX
  microsoft/onnxruntime#33992 · Filter by author JustinTArthur opened on Nov 8, 2026

• Fix duplicated word in build info comment

  microsoft/onnxruntime#33989 · Filter by author ddlw opened on Nov 8, 2026

• Symbolic shape inference gives SoftmaxCrossEntropyLoss a scalar loss when reduction is none

  microsoft/onnxruntime#33984 · Filter by author Shizoqua opened on Nov 8, 2026

• Symbolic shape inference gives LayerNormalization Mean/InvStdDev the input dtype instead of stash_type

  microsoft/onnxruntime#33982 · Filter by author Shizoqua opened on Nov 8, 2026

• onnx_test_data_utils image_to_pb --resize fails with Pillow 10+ and crops off by one pixel

  microsoft/onnxruntime#33980 · Filter by author sohailshk opened on Nov 8, 2026

• ReluQuantRewrite removes Relu without reading y_scale, changing outputs when scale is non-positive

  quantization
  microsoft/onnxruntime#33977 · Filter by author diansen33 opened on Nov 8, 2026

• [Mobile] onnxruntime-react-native has not been published to npm since 1.24.3

  platform:mobileplatform:web
  microsoft/onnxruntime#33971 · Filter by author joeytheman opened on Nov 7, 2026

• [Web] WebGPU: 4-bit MatMulNBits model ~80× slower than fp16 on an Imagination GPU (Pixel 10 Pro)

  ep:WebGPUplatform:mobileplatform:web
  microsoft/onnxruntime#33970 · Filter by author ToonTalk opened on Nov 7, 2026

• [Web] WebGPU builds have no CPU kernel for GatherBlockQuantized: reduced-type filter drops every two-typed registration

  api:Javascriptep:WebGPUplatform:webquantization
  microsoft/onnxruntime#33961 · Filter by author daveycodez opened on Nov 7, 2026

• External data from the Hugging Face cache fails to load with huggingface_hub >= 1.32 (shared blob store): symlink fallback from #27374 assumes one blobs/ folder

  microsoft/onnxruntime#33950 · Filter by author prometheus-maker opened on Nov 6, 2026

• [Performance] CoreML MLProgram attention retains MPS arrays on macOS 27.0.1 (687-byte repro)

  ep:CoreMLplatform:mobile
  microsoft/onnxruntime#33926 · Filter by author davidnichols-ops opened on Nov 5, 2026
  3

• [Bug] OrtGraphToProto rejects FLOAT16 tensor attributes (dtype 10), reproducible with ConstantOfShape

  microsoft/onnxruntime#33919 · Filter by author DingmaomaoBJTU opened on Nov 5, 2026
  1

• [Performance] ReduceLogSumExp (CPU EP, float, empty axes with noop_with_empty_axes=1) is ~36x slower since 1.24.1 (#26436)

  performance
  microsoft/onnxruntime#33910 · Filter by author junghyunpark2001 opened on Nov 5, 2026

• CUDA provider has no bool ReduceMax kernel

  ep:CUDA
  microsoft/onnxruntime#33876 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026

• CUDA graph replay of a graph with GatherND fails with an illegal memory access

  ep:CUDA
  microsoft/onnxruntime#33875 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026
  

• Pad has no bfloat16 kernel (CPU or CUDA); session fails to initialize with an unclear error

  ep:CUDA
  microsoft/onnxruntime#33874 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026

• [Performance] CPU MultiHeadAttention with attention_bias is ~4x slower than the unfused ops

  model:transformer
  microsoft/onnxruntime#33873 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026
  

• CUDA GroupQueryAttention with an attention_bias runs an unfused kernel, 4x slower than MultiHeadAttention with the same bias

  ep:CUDA
  microsoft/onnxruntime#33869 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026

• CPU MoE gives wrong results when its input is computed in the graph and its output feeds another node

  microsoft/onnxruntime#33866 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026
  1

  

• GroupQueryAttention fails to load when total_sequence_length is computed from static shapes (ParseData type mismatch)

  microsoft/onnxruntime#33865 · Filter by author IlyasMoutawwakil opened on Nov 4, 2026