Qwen 4 announces itself, Hydrogen-1.0 is there to download: the week of Chinese models

On November 6, 2026, at noon UTC, Qwen published its first official teaser; no weights online. On November 2, Deep Principle released Hydrogen-1.0, a GLM-5.3 post-trained on its lab notebooks, under the MIT license, claiming to beat GPT-6 Astra on FrontierScience-Research. And GLM-5.3 is running at T-Systems in Munich.

In short

• Qwen 4 is not out: on November 6, 2026 at 12:00 UTC, Qwen published its first official teaser, a flock of birds forming a fish and "The becoming begins". Nothing on Hugging Face or GitHub.
• On November 2, Deep Principle, in Hangzhou, published Hydrogen-1.0: GLM-5.3, 753 billion parameters, post-trained on chemistry, materials and its own lab notebooks. MIT license, 19,974 downloads on November 6.
• The publisher's scores, not replayed: 38.3% versus 33.3% for GPT-6 Astra on FrontierScience-Research. Astra and Claude Opus 5.5 are excluded from the experiment planning evaluation for refusing more than half of the requests.
• GLM-5.3 has been running at T-Systems in Munich since November 2, and Mistral has been serving GLM 5.2 since August 11. Caveat: 753 billion parameters do not run on a desktop machine.

← All the watch