---
format: "aidr-story-markdown/v1"
id: "79efbb054d4af9fc0fe98fd8d4b5c3ece82fe6c5c8ebc8cd86ca7158511b8b66"
canonical_url: "https://aidr.today/79efbb05?lang=en"
title: "Qwen 3.8 27B is excellent, but it defaults to overthinking things"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-08-16T23:45:09.000Z"
category: "Models"
topics: ["qwen","alibaba","llm","open-source"]
source_urls: ["https://simonwillison.net/2026/Aug/16/qwen-38-27b/","https://news.ycombinator.com/item?id=49324985"]
summary: "Friday’s big release was Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba’s Qwen research lab. I’ve been looking forward to this one: 27B is an …"
---

# Qwen 3\.8 27B is excellent, but it defaults to overthinking things

> [Open the canonical story](<https://aidr.today/79efbb05?lang=en>)

**Published:** 2026-08-16T23:45:09.000Z
**Category:** Models
**Topics:** qwen, alibaba, llm, open\-source

## Summary

Friday’s big release was Qwen 3\.8 27B, an Apache 2 licensed 27B parameter vision\-capable LLM from Alibaba’s Qwen research lab\. I’ve been looking forward to this one: 27B is an …

## Sources

- [Story source](<https://simonwillison.net/2026/Aug/16/qwen-38-27b/>)
- [Discussion](<https://news.ycombinator.com/item?id=49324985>)