Microsoft introduced seven in-house artificial intelligence models on Wednesday, asserting that its flagship reasoning and image-generation systems beat comparable offerings from Anthropic, OpenAI, and Google. The announcement came without third-party benchmarks, leaving the claims largely unverified.
The seven new models
Microsoft said the models cover a range of tasks, including language processing, code generation, and multimodal analysis. The company did not name each model individually but described them as part of an internal push to reduce reliance on external providers while competing directly with the AI leaders.
Flagship reasoning and image systems
Two models stood out in the announcement: a reasoning system designed to handle complex logic and mathematics, and an image-generation system that can create and edit visuals. Microsoft claimed both outperform comparable systems from Anthropic, OpenAI, and Google in internal tests. The company offered no details on the testing methodology or sample results.
Competitive claims
The move positions Microsoft as a direct rival to the AI labs it has invested in and partnered with. The company has poured billions into OpenAI and uses that startup’s models in its Copilot products. With these seven new models, Microsoft is now building its own competitive stack, a shift that could reshape its relationships with those partners.
Anthropic, OpenAI, and Google have not publicly responded to Microsoft’s claims.
What remains unclear
Microsoft has not released any independent evaluation of the models’ performance. Without external validation, the accuracy of the outperformance claims is unknown. The company said it plans to integrate the new models into its Azure AI platform and some consumer products, but gave no timeline.
Developers and enterprise customers will have to wait for benchmark results or public availability before testing Microsoft’s claims against competing systems themselves.
