Microsoft releases seven in-house models at Ignite, replacing external AI models in its products
Microsoft is replacing outside AI models in its own products with cheaper in-house alternatives
AI summary
Microsoft releases seven in-house models at Ignite, replacing external models in its own products. Microsoft says MAI Code 2 averages 98% of GPT-6.1 Sol on four benchmarks at only 40% of the cost, and 96% of the performance of Claude Opus 5.5 at 5% of the cost. MAI Cyber 2 Flash solves 92.3% of tasks on CyberGym-E2E, above GPT-6 Astra's 91.5%.