0.16.0

Issues

• [discussions]: [Compatibility/vLLM] AutoScheme mixed-bit export (llm_compressor) fails in vLLM with AssertionError in load_merged_column_weight

  bug
  intel/auto-round#2174 · Filter by author vishvaRam opened on Aug 17, 2026 · 0.16.0
  1

  

• [Bug]: woq_gemm's m > 1 path submits its oneDNN matmul to a cached stream bound to the first queue ever seen, so it is never recorded into a torch.xpu graph and every replay returns zeros

  intel/auto-round#2206 · Filter by author alankessler opened on Aug 21, 2026 · 0.16.0
  5

  

• [Bug]: Fix AutoRound gradients for Kimi-VL routed experts- #2097

  bug
  intel/auto-round#2282 · Filter by author wenhuach21 opened on Sep 3, 2026 · 0.16.0
  

• [Upstream]: vLLM inference with AR kernels

  high priority
  intel/auto-round#1446 · Filter by author luoyu-intel opened on Feb 12, 2026 · 0.16.0
  1

  

• [Bug]: MiniMax-M2.7_mxfp4_mixed_fp8kv_LLMC vllm eval failed

  bug
  Bug

  intel/auto-round#2369 · Filter by author chensuyue opened on Sep 16, 2026 · 0.16.0
  

• [Bug]: Kimi-K2.6_mxfp4_fp8attn_LLMC vllm eval qkv scale is not loaded

  bug
  Bug

  intel/auto-round#2370 · Filter by author chensuyue opened on Sep 16, 2026 · 0.16.0
  2

  

• [gemma-4-31B] 4-GPUs RuntimeError: cannot enter context: <_contextvars.Context object at 0x7e35759202c0> is already entered

  intel/auto-round#2410 · Filter by author XuehaoSun opened on Sep 23, 2026 · 0.16.0
  

• [llama3.1 8b][mxfp8] inference error on sm120 when using FLASHINFER attention backend

  intel/auto-round#2413 · Filter by author XuehaoSun opened on Sep 23, 2026 · 0.16.0
  1