Quasar 438B is Multiverse Computing's flagship reasoning model, its first in the 400B-plus class, built for enterprise agents and coding. It runs in English and Spanish. On the four published charts: Terminal-Bench v2.1 at 69.3, AA-LCR long-context at 75.0, 500 output tokens in 15.3 seconds including thinking time, and a 43 composite Artificial Analysis Intelligence Index. It's live via the CompactifAI API.

Europe's leading model now comes from a company whose core business is quantum computing, not LLM pretraining — the same efficiency thesis that let a 438B model answer in 15 seconds. Whoever controls cost and latency can enter the frontier race late and still take regional leadership.

Quasar 438B: Europe's Top AI Model at Index 43

Multiverse Computing — a quantum-software shop, not an LLM lab — released Quasar 438B today, and it now holds the highest Artificial Analysis Intelligence Index score of any European model: 43 on v4.1.1. That puts it 13 points ahead of Mistral Medium 3.5 (30) and above Nemotron 3 Ultra (38) and Inkling (42). The crown quietly changed hands.

What shipped

Quasar 438B is Multiverse's flagship reasoning model, its first in the 400B-plus class, built for enterprise agents and coding. It runs in English and Spanish. On the four published charts: Terminal-Bench v2.1 at 69.3 (beating Mistral Medium 3.5 by 18.7 points, within striking range of Claude Opus 5's 89.1), AA-LCR long-context at 75.0 (level with Grok 4.6, within a point of Claude Opus 5), 500 output tokens in 15.3 seconds including thinking time — only Gemini 3.7 Flash is both faster and smarter — and a 43 composite. It's live via the CompactifAI API, no infrastructure required.

Why it matters

The pattern isn't the benchmark. It's who shipped it. Europe's leading model now comes from a company whose core business is quantum computing, not LLM pretraining — the same way the efficiency thesis (CompactifAI) that made them money on inference is what let a 438B model answer in 15 seconds. The efficient-frontier story keeps repeating: whoever controls cost and latency can enter the frontier race late and still take regional leadership. Watch whether CompactifAI pricing undercuts the usual 400B-class API rates — that's the real tell.

Verdict

Quasar 438B is the strongest signal out of European AI this year: a new regional leader, a quantum shop as the surprise source, and speed that undercuts the parameter class. If you build agents on European models, this changes the shortlist.