Issues

• Add Qwen3.8‑Flash‑Next ParoQuant

  z-lab/paroquant#55 · Filter by author GGZ14 opened on Sep 3, 2026
  1

• [Performance][vLLM] Need profiling guidance for ParoQuant runtime overhead vs AWQ-Marlin on Qwen3.5-0.8B/27B

  z-lab/paroquant#52 · Filter by author Jealousc11gx opened on Jun 15, 2026

• Text-only PARO MoE checkpoints ship a vision_config → servers mis-route to VLM engine (load failure / SIGKILL)

  z-lab/paroquant#49 · Filter by author sangemaru opened on May 22, 2026
  1

• Qwen3.6-35B-A3B-PARO: config.json declares mtp_num_hidden_layers=1 but no MTP weights shipped → 0% accept + ~50% decode overhead by default

  z-lab/paroquant#47 · Filter by author sangemaru opened on May 19, 2026
  2

• when will we upload qwen3-32b-paro model in huggingface?

  z-lab/paroquant#45 · Filter by author dongdongvs6 opened on May 18, 2026
  4

• Request Jackrong/Qwopus3.5-27B-v3 Paro version

  model request
  z-lab/paroquant#27 · Filter by author menciuschow opened on Apr 5, 2026

• rocm support

  z-lab/paroquant#25 · Filter by author ambyjkl opened on Mar 29, 2026
  1

• Can't disable thinking/reasoning mode + JSON format issues with OpenAI-compatible API

  z-lab/paroquant#24 · Filter by author Milor123 opened on Mar 23, 2026
  1

• Model Requests

  model request
  z-lab/paroquant#23 · Filter by author copilotprosub1 opened on Mar 22, 2026

• Request for 2-bit, 3-bit, and 4-bit quantized versions of LLaMA 3.1-70B and LLaMA 2-70B

  model request
  z-lab/paroquant#22 · Filter by author Hittopu opened on Mar 20, 2026
  2

• Empty models list returned from /v1/models endpoint

  z-lab/paroquant#20 · Filter by author zlatent opened on Mar 19, 2026

• can you have a benchmark on unsloth UD2.0 with the same quant bit

  z-lab/paroquant#13 · Filter by author cresseelia opened on Mar 11, 2026