Microsoft Unveils Six Specialized AI Models to Challenge OpenAI and Anthropic

Microsoft introduced six new specialized AI models at its Ignite conference in San Francisco, targeting specific tasks such as coding, cybersecurity, image generation, and voice. The company aims to compete with industry leaders by offering performance comparable to frontier models at a significantly lower cost. These new offerings include MAI Code 2, which Microsoft claims delivers 98% of GPT-6.1 Sol's performance at 40% of the price and 96% of Anthropic's Claude Opus 5.5 performance at 5% of the cost.

Other releases include MAI Code 2 Flash, a faster model that Microsoft says matches GPT-6 Luna at 65% of the price and Claude Haiku 5.5 at 20% of the cost. The company also debuted MAI Cyber 2 Flash, a cybersecurity model that reportedly beats GPT-6 Astra on security benchmarks at less than 2% of the cost per task. Additional models include MAI Image 2.7, which Microsoft says ranks first in image editing, alongside new transcription and real-time voice models.

Microsoft AI CEO Mustafa Suleyman told Axios that the company is making significant progress toward building its own frontier general-purpose AI model. While the company is focusing on specialized models to reduce costs, Suleyman noted that the strategy involves packing more expertise into the same number of parameters. "It basically means that you cram more tokens and knowledge and expertise into the same amount of parameters," Suleyman said.

The company is also offering enterprise customers access to licensed model weights, allowing for customization with proprietary data within the Azure environment. Suleyman stated this approach would "feel like open source" while maintaining security and control. The move could potentially shift workloads away from outside providers, as MAI Code 2 is slated for availability in GitHub Copilot.

Microsoft's performance comparisons rely on selected benchmarks, and some pricing advantages reflect promotional discounts. The company's push for specialized models comes as OpenAI announced new models, including GPT-6.2 Astra and GPT-6.2 Sol, with the Astra model priced 20 percent less than its predecessor. Suleyman confirmed Microsoft has achieved state-of-the-art pre-training results and is making progress in reinforcement learning, though he declined to provide a timeline for a general-purpose model release.

"My team is building true superintelligence," Suleyman said.

FI Journal
Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.