EU-AI Monthly Update #1: The rise of Chinese Models, A "new" 438B model and Domyn raises over $1bn.

The key EU-AI events, developments & controversies:

• 🇮🇹 Milan’s Domyn raises over $1bn in Europe’s largest recent AI round.
• 🇫🇷 Mistral releases GLM 5.2 (Chinese Model by z.ai) on their EU-hosted infrastructure as alternative to Mistral Medium.
• 🇪🇸 Multiverse Computing Introduces Quasar 438B: Europe's Leading AI Model - later confirmed as compressed GLM 5.2.
• Many are skeptical of the rise and use of Chinese Models hosted on EU-infrastructure, but they do live up to their benchmarks + are more cost efficient.

---

5 september 2026 - Written by M.J (not AI assisted):

A lot has happened in just one month, and it's interesting that the first post on this website is about the rise of chinese models.

Nevertheless, overall we are making progress.

Mistral is still working on "their new model", probably the Mistral 4 family.

Milan's Domyn raises over $1bn, placing them in a similar league as the previous crazy $1bn raise by AMI Labs in terms of amounts of money raised in relation to EU-AI.

And we're even seeing smaller labs achieve better performance for the same amount of parameters with experimential new methods like:

• 🇩🇰 Danish Foundation Models releases Mimir v1 (1B HRM-Text, outperforming similar small models)

The biggest new mover: Multiverse Computing.

One of the biggest new mover we've seen so far is by Multiverse Computing, releasing their "new" 438B model Quasar that is beating Mistral Medium on many benchmarks.

Figure 1. Benchmark score (Terminal-Bench v2.1). Higher is better. Source: Multiverse Computing.

---

Many were sceptical of these claims, and later though, they confirmed that the Quasar 438B model is in fact a "compressed" version of GLM 5.2.

Be that as it may, they are currently offering this compressed high performing model for a significantly better price per million input/output tokens than any of their competitors, including Mistral.

What's next? And what's the optimal model now?

Will we see a rise of more Chinese model providers? Are we in a race to bottom? Or are there new (perhaps currently stealth) foundational AI companies and projects on the horizon?

Only time will tell.

For the short-term, it seems Europe's most optimal and cost-efficient AI systems will run on GLM.

Whether that's the full GLM model, e.g. offered by Mistral or other flavors of GLM, like the compressed GLM offered by Multiverse Computing.