// AI NEWS · MODEL RELEASE

Multiverse Computing Launches Quasar 438B, Europe's Highest-Scoring Model on the Artificial Analysis Intelligence Index

The San Sebastián company's first large model is a 438-billion-parameter reasoning model built for agents and coding, running in English and Spanish — pitched as European sovereign AI that competes on capability and speed.

2026-09-02 · by Moe Ameen

What happened

Multiverse Computing released Quasar 438B on September 2, 2026 — its first large model and, according to independent evaluator Artificial Analysis, the highest-scoring European model on the Intelligence Index. Multiverse is a San Sebastián, Spain company best known for its work compressing large AI models via its CompactifAI platform; Quasar is a 438-billion-parameter reasoning model aimed at enterprise-scale agents, coding, and multi-step reasoning, and it operates in English and Spanish.

The headline number is a score of 43 on the Artificial Analysis Intelligence Index v4.1.1. That places it ahead of other European entries and models like Mistral Medium 3.5 (30) and NVIDIA Nemotron 3 Ultra (38), while still trailing frontier US models such as Claude Opus 5 (63). Multiverse pairs the score with speed: the model returns 500 output tokens in about 15.3 seconds, and posts 69.3 on Terminal-Bench v2.1 (agentic terminal work) and 75.0 on AA-LCR (long-context reasoning). CEO Enrique Lizaso framed the launch around not forcing European developers to choose between reasoning performance and speed.

The positioning is European sovereign AI — a domestically built model that enterprises in Europe can adopt without depending on US or Chinese providers. Quasar is available for testing through Multiverse's CompactifAI API, with sign-up at dashboard.compactif.ai, and the company points organizations to business@multiversecomputing.com. Multiverse did not publish per-token pricing at launch. Treat the exact scores, languages, and availability as launch-day details and confirm them on Multiverse's own materials.

Why it matters for creators

  • A fast, native English-and-Spanish model matters most to bilingual and Spanish-language creators. Quasar was built for both languages rather than being English-first with Spanish bolted on, which is rarer at this capability tier than it should be.
  • It is built for agents and coding, not creator content. The benchmarks that moved — Terminal-Bench, long-context reasoning — are enterprise-automation scores, not "writes a scroll-stopping hook" scores. A high Intelligence Index rank does not make a model a content tool.
  • European sovereign AI is becoming a real procurement axis. Agencies and creators serving EU brands with data-residency requirements now have a domestic, high-scoring option to draft with, which can matter more than a leaderboard rank in a pitch.
  • Speed is the quiet story. Returning 500 tokens in roughly 15 seconds is quick for a 438B model, and for interactive drafting that responsiveness matters far more day to day than a single index number.
  • Access is API-only right now. There is no consumer app, no image or video generation, and no scheduling — Quasar is a raw model you or your tools call, so the work of turning a draft into finished, published content is still entirely on you.

How to act on this with Kompozy

You cannot publish a leaderboard score. What actually shipped on September 2 is a strong, fast, bilingual text model — a better drafting engine, not a content operation. The gap between "Quasar wrote me good English and Spanish copy" and "that copy is live as a captioned video, a branded carousel, a blog, and a newsletter across all my platforms" is exactly the gap [Kompozy](/) closes.

Here is the move a creator makes with this news today. Draft your angle in Quasar — a script, a thread, a long-form outline in either English or Spanish — then bring that text into Kompozy, where one idea fans into 18 formats: [Persona Shorts](/glossary/persona-shorts) and HeyGen avatar clips that put the script on camera, brand-exact [Carousels](/glossary/hyperframes) through HyperFrames, quote graphics, a blog article, and a matching newsletter. Because Quasar is bilingual, this is where sovereignty gets practical: run one [Persona Brief](/glossary/persona-brief) for your English audience and a second for your Spanish one, and Kompozy keeps each batch reading as one consistent creator rather than a translation. Then [Autopilot](/glossary/autopilot) schedules and publishes the whole set across the eight social platforms plus blog and email behind a per-post review. Quasar is the drafting brain; Kompozy is the studio and the distribution that a model, by itself, will never be.

Quick takeaways

  • Multiverse Computing launched Quasar 438B on September 2, 2026 — its first large model and, per Artificial Analysis, the highest-scoring European model on the Intelligence Index (score 43 on v4.1.1).
  • It is a 438-billion-parameter reasoning model for enterprise agents, coding, and long-context reasoning, operating in English and Spanish.
  • Benchmarks: 43 on the Intelligence Index (ahead of Mistral Medium 3.5 and NVIDIA Nemotron 3 Ultra, behind Claude Opus 5 at 63), 69.3 on Terminal-Bench v2.1, 75.0 on AA-LCR, and ~15.3s to return 500 tokens.
  • Positioned as European sovereign AI; available via the CompactifAI API (dashboard.compactif.ai). Per-token pricing was not disclosed at launch.
  • It is an API-only text model — no image, video, or publishing — so turning a draft into finished, on-brand content across platforms is still a separate step, which is where a content engine like Kompozy fits.

Frequently asked questions

What is Quasar 438B?

Quasar 438B is Multiverse Computing's first large model, launched September 2, 2026. It is a 438-billion-parameter reasoning model built for enterprise-scale agents, coding, and multi-step reasoning, operating in English and Spanish, and delivered through the CompactifAI API. Artificial Analysis rated it the highest-scoring European model on its Intelligence Index.

How does Quasar 438B compare to other AI models?

On the Artificial Analysis Intelligence Index v4.1.1 it scores 43 — the top European result, ahead of Mistral Medium 3.5 (30) and NVIDIA Nemotron 3 Ultra (38), but behind frontier US models such as Claude Opus 5 (63). It also posts 69.3 on Terminal-Bench v2.1 and 75.0 on AA-LCR, and returns 500 tokens in about 15.3 seconds, which is fast for its size.

Can I use Quasar 438B to create social content?

It can draft text in English and Spanish, but it is an API-only reasoning model with no image or video generation and no publishing. To turn Quasar drafts into finished posts — captioned video, carousels, blogs, newsletters — and schedule them across platforms, you pair it with a content engine like Kompozy that handles the formats and distribution.

Where can I access Quasar 438B?

Quasar is available for testing through Multiverse Computing's CompactifAI API, with sign-up at dashboard.compactif.ai, and organizations can contact business@multiversecomputing.com. Multiverse did not publish per-token pricing at launch, so confirm current access and cost on its official pages.

Related news

← All AI news · Get started →