Tencent's Hy4 Preview defaults to high reasoning when its supplied chat template receives no reasoning setting. As Simon Willison noticed in his inspection of the release, the template accepts only high and no_think, rather than a range of intermediate levels.
The published template confirms that behavior and rejects other values. Applications using it therefore need to choose no_think explicitly when they want reasoning disabled. This is a concrete integration detail revealed by the model's input-formatting code, without requiring a judgment about its overall benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → performance.
Terms in this piece · Glossary
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
