Qwen 3.8 27B: Great Benchmarks, Terrible Habit of Overthinking Everything
Qwen 3.8 27B, an Apache 2.0 licensed vision-capable model from Alibaba, scores 52 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Luna and trailing GLM-5.2 and DeepSeek V4 Pro by one point despite being far smaller.
What it is
Qwen 3.8 27B is Alibaba's Qwen research lab's newest openly licensed vision-capable LLM, sized to run on a well-specced laptop.
What it does
It scores 52 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Luna and landing one point behind GLM-5.2 (753B parameters) and DeepSeek V4 Pro 0813 (1.7 trillion parameters), both vastly larger models. Simon Willison notes it is excellent but defaults to wildly overthinking straightforward requests, generating far more reasoning than the task needs.
Why it matters
A 27B model scoring within a point of models 20 to 60 times its size is a genuine efficiency story, but the overthinking tendency means raw benchmark scores don't tell you what it'll cost you in tokens and latency for everyday use, an alternative to running one of the much larger models locally if you can tolerate tuning its reasoning budget.
How to use it
Download Qwen 3.8 27B from Hugging Face and run it locally via LM Studio or a compatible inference server; expect to tune reasoning effort to control overthinking.
Tencent's Hy4 Preview: 770B Params, No Vision, 1M Context
Hy4 Preview is Tencent's new open-weight, text-only large language model with 770B total parameters, 49B active parameters, and a 1M token context window.
llm-anthropic 0.27: Compatibility Patch for the New anthropic v1.0.0 SDK
llm-anthropic 0.27 updates Simon Willison's Anthropic plugin for LLM to work with anthropic v1.0.0, which switched its HTTP client from httpx to httpx2.
Graham Dumpleton Is Writing a Wrapture Tutorial Almost Every Day
Graham Dumpleton has been publishing near-daily tutorials for wrapture, his Python function-wrapping library, since its August 31st release, expanding beyond the original announcement.