Microsoft launched seven in-house MAI AI models at its November 17 Ignite conference, spanning coding, cybersecurity, imaging, and real-time voice applications. The company claims MAI Code 2 achieves approximately 98% of GPT-6.1 Sol's performance on enterprise coding benchmarks at roughly 40% of the cost, and about 96% of Claude Opus 5.5's performance at approximately 5% of its price.
MAI Cyber 2 Flash reportedly matches GPT-6 Astra in quality on the CyberGymE2E benchmark at less than 2% of per-task cost, though some comparisons use promotional pricing. MAI Code 2 and Code 2 Flash are now available on GitHub Copilot, with Microsoft also offering licensed model weights for enterprises to customize and deploy on Azure. The release coincides with OpenAI's official launch of GPT-6.2 Astra and Sol on the same day.
Microsoft Releases Seven In-House MAI Models, Claims Near Parity With OpenAI and Anthropic at Fractional Cost
免責事項: Phemexニュースで提供されるコンテンツは、あくまで情報提供を目的としたものであり、第三者の記事から取得した情報の正確性・完全性・信頼性について保証するものではありません。本コンテンツは金融または投資の助言を目的としたものではなく、投資に関する最終判断はご自身での調査と、信頼できる専門家への相談を踏まえて行ってください。
