Qwen3.8-Flash-Next Stuck in Endless Thinking at Low Reasoning Level
pabloodiablo · reddit · 2026-08-30
A developer reported severe performance issues with Qwen3.8-Flash-Next (Q8 quantized) on a 2x 128GB VRAM cluster. Even with the reasoning effort parameter set to low, the model entered an "endless thinking" loop during code generation, running for over 45 minutes without stopping (reaching 32k context), while generating significantly more output than comparable models (Qwen3.8-27B and DeepSeek v4 both finished in under 15 minutes). The author suspects a llama.cpp configuration issue or a bug in the model and is seeking community input.
More from Models
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Argon edges out astra as daily driver, but Claude still wins heavy coding work — iruletheworldmo · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- RoboVista benchmark: Gemini-4 Argon hits 81.9% on robotics understanding — berkeley_ai · 2026-10-16
- Dev reports decision model jev beats gpt-6-luna on quality in real training pipeline — genmon · 2026-10-16
- Google won't publish Argon 4 limits in AI Pro—users compare experiences — Consistent-Text-2662 · 2026-10-16