Spanish AI company unveils Quasar 438B, matching Qwen 2 7B in low-inference benchmarks to become Europe’s top AI model.

Insight Beating AI News: Spain-based Multiverse Computing has officially launched Quasar 438B, a 438-billion-parameter inference model. Independent testing by Artificial Analysis awarded it a score of 43, outperforming Mistral Medium 3.5’s 30 points to become the highest-scoring European model on the current leaderboard. That said, this “European No.1” lags far behind global front-runners: in Artificial Analysis’s AI Index, Claude Fable 5.1 Max’s inference tier scores 66 points. Alibaba’s Qwen3.8 27B, with just 27 billion parameters, matches Quasar’s 43 points in the low inference tier, hits 44 in medium, and reaches 52 in xhigh. Quasar’s weights are not open, and Multiverse has not disclosed its model architecture, training data, or actual activated parameter count. Some community members have questioned whether it is fine-tuned from other models, but official documentation explicitly classifies Quasar as an “Original Model”.