Multiverse Computing launches Quasar 438B, the highest-scoring European model on Artificial Analysis Intelligence Index

The launch represents a significant step for European sovereign AI, demonstrating that Europe can compete with leading models from the US and China on capability and speed.

The new model:

• Outperforms Mistral Medium 3.5, which scores 30, and NVIDIA Nemotron 3 Ultra, which scores 38.
• Produces 500 output tokens in 15.3 seconds, including reasoning time. Only three models in the comparison are faster, and only one of them — Gemini 3.7 Flash — also records a higher Intelligence Index score. Quasar is faster than Mistral Medium 3.5, which takes 18.8 seconds, while delivering a 13-point higher score.

That difference becomes particularly important in agentic systems. A single user request may require a model to plan a task, call several tools, check the results and adjust its approach. When an agent makes dozens of model calls to complete one piece of work, delays at each stage compound into minutes of additional waiting time. Quasar is designed to keep those loops moving while retaining the reasoning ability needed for complex work.

The launch takes Multiverse Computing’s work on efficient, deployable AI into the 400-billion-plus parameter class. Quasar has 438 billion parameters, giving it the scale needed for demanding reasoning tasks while addressing the latency that can make very large models difficult to use inside interactive enterprise products.

“This is a significant milestone for European AI: Quasar shows that European AI developers do not have to choose between reasoning performance and speed,” said Enrique Lizaso, co-founder and CEO of Multiverse Computing. “European enterprises need models that can work through complex tasks, use tools and handle long documents, and they also need greater choice and access to powerful AI developed here in Europe. Quasar brings those two requirements together in a model built for enterprise agents and coding.”

Strong performance across long-context reasoning and coding

Quasar scores 75.0 on Artificial Analysis Long Context Reasoning (AA-LCR), which tests a model’s ability to extract, connect and reason over information distributed across long documents. The result matches Grok 4.6 (high), comes within one point of Claude Opus 5 and leads Nemotron 3 Ultra by 4.0 points and Mistral Medium 3.5 by 9.7 points.

On Terminal-Bench v2.1, which evaluates agents working in real terminal environments, Quasar scores 69.3. That places it 18.7 points ahead of Mistral Medium 3.5 and 15.4 points ahead of Nemotron 3 Ultra.

The Terminal-Bench result reflects Quasar’s ability to work through practical coding and command-line tasks rather than simply generate isolated snippets of code. These capabilities are relevant to software-development agents that need to inspect repositories, run commands, diagnose errors and complete a sequence of connected actions. Quasar trails the frontier result on this evaluation, led by Claude Opus 5 at 89.1, leaving a clear area for further development.

Its long-context performance is particularly relevant to enterprise research and knowledge work. Many business tasks depend on information spread across contracts, technical documentation, policies, reports or multiple earlier interactions. A model must be able to find the relevant details, connect them and reason across the full body of material rather than treat each passage in isolation.

Artificial Analysis Intelligence Index v4.1.1 is a weighted average of nine evaluations across four categories: Agents (GDPval-AA v2 and τ³-Banking), Coding (Terminal-Bench v2.1 and SciCode), Scientific Reasoning (Humanity's Last Exam, GPQA Diamond and CritPt) and General knowledge and long-context reasoning (AA-Omniscience and AA-LCR). Quasar outperforms comparable European models on eight of the nine evaluations.

Available through the CompactifAI API

Quasar 438B is available now through the CompactifAI API, allowing organisations to test and integrate the model without setting up their own infrastructure.

Support for both English and Spanish makes the model suitable for European and international organisations that need to deploy the same reasoning system across teams and markets. Potential applications include software engineering, operational automation, document-heavy research and technical knowledge work, where accuracy, context retention and completion time all affect whether an AI system is useful in practice.

Multiverse Computing will continue to develop Quasar’s coding and agentic capabilities following the launch, with further updates planned for the model.

To access the model, please contact: business@multiversecomputing.com

About Multiverse Computing

Multiverse Computing is a leader in sovereign and efficient AI. The company develops fast, efficient, and highly specialised AI models that enable organizations to deploy advanced artificial intelligence securely within their own infrastructure, ensuring full control over data, governance, and compliance. Headquartered in Donostia-San Sebastián, Spain, with offices in the United States, Canada, and across Europe, Multiverse serves more than 100 global customers, including Iberdrola, Bosch, and the Bank of Canada.

For more information, visit multiversecomputing.com.

Media Contacts:US/North AmericaLaunchSquad for Multiverse Computingmultiverse@launchsquad.com

EuropeBallou PR for Multiverse Computingmultiversecomputing@balloupr.com

A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/40f147b1-ed9b-4279-a22b-c68ec74b1776

NEWS RELEASE TRANSMITTED BY Globe Newswire