Published 2 days ago • loading... • Updated 23 hours ago
Microsoft Unveils AI Models in Push for Independence From OpenAI
The models are trained from scratch on licensed data as Microsoft seeks to cut fees and reduce reliance on outside AI providers.
Microsoft unveiled seven in-house AI models at the Build conference in San Francisco, signaling a strategic shift to reduce reliance on third-party partners like OpenAI and Anthropic.
Joining the broader 'agentic' AI industry trend, Microsoft launched Microsoft Scout to automate meeting prep and email drafting; rival Google released Gemini Spark for premium subscribers last week.
The flagship MAI-Thinking-1 model features 35 billion active parameters and a 256,000-token context window, built 'from scratch' with 'no distillation' to outperform Claude Sonnet 4.6 on coding benchmarks.
Microsoft CEO Satya Nadella said, "We believe the time has come for every company to move from consuming a frontier model to fully participating at the frontier," as Anthropic prepares for an IPO.
Corporate Vice President Zulfi Alam said, "We will have a quantum machine in 2029 that can solve commercially viable, reasonable problems," as Microsoft Quantum announced the Majorana 2 chip is 1,000 times more reliable than its predecessor.
Redmond's firm announced on Tuesday seven models, in order to be less dependent, in artificial intelligence, on its partnership with ChatGPT's parent company.