Spain's quasar 438b model leads european AI benchmark with 1m context window

Multiverse Computing, a Spanish AI firm, has launched the Quasar 438B model, which ranks highest among European models in the Intelligence Index v4.1.1 with a score of 43. The model, with 4.38 trillion parameters, is designed for enterprise use, software development, and complex multi-step tasks. It supports English and Spanish and is available via the CompactifAI API. The model’s context window spans 1 million tokens, making it suitable for handling long texts like contracts and technical documents. Quasar 438B achieves a score of 75.0 in long-context reasoning tests and 69.3 in Terminal-Bench v2.1, simulating real-world terminal tasks. The model is proprietary, with weights not publicly available, and offers API pricing at $0.60 per million input tokens and $1.80 per million output tokens. According to the official X post, the model may be based on GLM5.2 from Zhipu AI.