Multiverse Computing Unveils 'Quasar 438B', Europe's Highest-Performing Inference Model

Multiverse Computing has released Quasar 438B, its flagship inference model designed for enterprise-scale agents and coding. The model scored 43 points on the Artificial Analysis Intelligence Index v4.1.1, achieving the highest score among European models and surpassing Mistral Medium 3.5 (30 points) and NVIDIA Nemotron 3 Ultra (38 points).

Performance and Speed

Quasar 438B demonstrated competitiveness in both intelligence and speed. It takes 15.3 seconds to generate a 500-token response (including inference time), making it faster than the top three models among its comparators. Notably, among models with higher scores, none are faster than Quasar except for Gemini 3.7 Flash.

Key Benchmark Results

• Long-Context Reasoning (AA-LCR): Scored 75.0 points, tying with Grok 4.6 and performing nearly on par with Claude Opus 5 (75.7 points).
• Terminal-Based Coding (Terminal-Bench v2.1): Scored 69.3 points, which is 18.7 points higher than Mistral Medium 3.5 but lower than Claude Opus 5 (89.1 points).

Features and Use Cases

Quasar 438B is designed to minimize latency in the 400B+ parameter class, making it suitable for agentic workflows requiring multi-step tasks such as planning, tool use, and code execution. It supports English and Spanish and can be tested and deployed without infrastructure setup via the CompactifAI API.