Microsoft AI this month released three foundation models — a text model, a voice model, and a generative image model — marking the first time the company has shipped frontier-scale model weights of its own. For most of the last four years, Microsoft's AI products have been powered primarily by OpenAI models served through an exclusive Azure pipeline.
The three models — codenamed internally as 'Echo' (voice), 'Paint' (image), and 'Speak' (text) — were developed by the Microsoft AI organization under Mustafa Suleyman, which has grown to roughly 3,600 headcount since Suleyman joined in March 2024.
In Copilot, Microsoft is routing a portion of queries to the new Speak model and the rest to GPT-5.4. Internally, the company describes the split as 'capability-matched': OpenAI models are retained for the highest-bar reasoning and coding tasks, and Microsoft models handle the long tail of conversational and creative queries where cost and latency matter more than peak intelligence.
Key facts
- 01Microsoft. A key thread of reporting in this story.
- 02Copilot. A key thread of reporting in this story.
- 03Models. A key thread of reporting in this story.
The commercial implication is significant. Microsoft has been paying OpenAI substantial per-token fees inside its Azure commitment. A credible internal alternative lets Microsoft renegotiate that relationship from a much stronger position when the current agreement comes up for renewal in 2027.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
