7 November 2026

Several smaller AI model releases from Tencent, Nous, Qwen and others

First reported

Latent Space ran this on 7 November.

Tencent's Hy4 is reported as the top Chinese model on the Agent Arena leaderboard.
Nous Research released Arete-27B, which reportedly lowers how often the model produces propaganda.
Researchers showed that editing a model's internal weights can cut the number of thinking tokens it uses.

Our read

Only Latent Space of the 30 newsletters covered this, and the archive has a similar roundup of smaller model releases from 2026-10-20 that also included a fast OCR model. An earlier Qwen story from 2026-08-18 is the other item on this thread.

Earlier on this story

Oct 20Several smaller AI model releases land, including a fast OCR model
Aug 18Alibaba's smaller Qwen model matches larger competitors on benchmark

Written by Newskeryx from which newsletters covered this story and from our own archive.

How it was covered

Latent Spaceswyx & Alessio

Tencent Hy4 arena and cost results, Arete-27B values eval, Qwen weight-space edits, and Drex decision models