Qwen 3.8 27B is excellent, but it defaults to overthinking things
Alibaba's Qwen research lab recently released Qwen 3.8 27B, a vision-capable LLM with 27 billion parameters under the Apache 2 license. This model demonstrates significant improvements over its predecessor, Qwen 3.6 27B, and rivals like the closed-weight Qwen 3.7-Plus. In benchmark tests, Qwen 3.8 outperformed both of these models and even surpassed a larger Qwen 3.8 2.4T-A95B model.
When running Qwen 3.8 on a MacBook Pro and an NVIDIA DGX Spark, the model defaults to overthinking tasks due to its xhigh reasoning effort setting. This leads to longer processing times and excessive token usage, especially on consumer hardware. However, adjusting the reasoning_effort parameter can optimize the model's performance for specific tasks.
Despite its impressive capabilities, Qwen 3.8's default settings make it better suited for more complex tasks on powerful hardware rather than everyday use on laptops.
Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
