# Пов'язані статті щодо Health AI

Just Now, Zuckerberg's $14.3 Billion "Avocado" Is Here: Challenging GPT-5.4 Head-On, Silicon Valley's Most Expensive Chinese Debut

Meta has unveiled Muse Spark (codenamed Avocado), its first AI model from the Meta Superintelligence Lab (MSL), positioning it as a direct competitor to leading models like GPT-5.4 and Gemini 3.1 Pro. With a score of 52 on the Artificial Analysis benchmark, it significantly outperforms its predecessor, Llama 4 Maverick (18 points), and ranks among the top models. Key features include native multimodal perception, tool use, visual chain-of-thought reasoning, and multi-agent collaboration via a "Contemplating Mode." Muse Spark excels in multimodal tasks, health-related queries (trained with input from 1,000+ clinicians), and agent-based applications, though it lags slightly in coding and complex reasoning tasks. The model is highly token-efficient, using significantly fewer tokens than competitors like Opus 4.6 and GPT-5.4 for similar performance. It is integrated into Meta's platforms (meta.ai, Facebook, Instagram, WhatsApp) and is free to use, though it remains closed-source. The development involved a complete rebuild of Meta's AI stack, reducing pre-training compute needs to one-tenth of Llama 4's requirements. The project, led by Chief AI Officer Alexandr Wang and a team of researchers from OpenAI and DeepMind, marks Meta's re-entry into the top tier of AI development.

marsbit04/09 07:49