Ugh I didn't want to post this... Back to Qwen3.8 27B

A user expresses disappointment with Qwen Flash Next, despite its speed and occasional improvements over Qwen3.8 27B. They note that Flash Next at iq4_xs often fails to follow instructions, hallucinates more, and uses fewer tokens. While acknowledging its strength for one-shots and benchmarks, the user finds it less reliable for long-term agentic assistant work, contrasting it with the more consistent performance of Qwen3.8 27B. They seek others' experiences with Flash Next in agentic tasks and with different claw/hermes flavors.

This user review offers a direct comparison of Qwen Flash Next and Qwen3.8 27B, highlighting Flash Next's shortcomings in agentic work unlike its predecessor.