AnalysisAI ModelsSeptember 7, 2026

Qwen 3.8 Next Flash draws complaints over verbosity

A LocalLLaMA user reports 13 minutes of thinking time on single-turn coding requests with Qwen 3.8 Next Flash, at roughly 150 tokens/sec generation and 7,000 tokens/sec prompt processing. The poster switched from 3.6 27b, calling Next Flash a logical step up even from 3.8 27b.

1 source

More stories today

Open the live feed