Microsoft Fires Back at AI Rivals With Three New MAI Models — Cheaper Than Google and OpenAI

Microsoft's MAI Superintelligence team shipped three production-grade multimodal AI models to Microsoft Foundry today — and is pricing them below comparable offerings from Google and OpenAI.

The three new MAI models

• MAI-Transcribe-1 — Speech-to-text across 25 languages; 2.5× faster than Azure Fast; starts at $0.36/hour
• MAI-Voice-1 — Audio generation; produces 60 seconds of voice in 1 second; custom voice support; starts at $22/million characters
• MAI-Image-2 — Image/video generation (originally launched March 19 on MAI Playground); starts at $5/image

Why it matters

Microsoft is actively building its own model stack alongside its OpenAI partnership. A recently renegotiated contract now explicitly allows Microsoft to pursue its own superintelligence research — giving Mustafa Suleyman's MAI team the green light to compete directly with OpenAI and Google on foundation models.

Suleyman in the announcement blog post: "You'll see more models from us soon in Foundry and directly in Microsoft products and experiences."

Context: just this week, Microsoft also launched Copilot Cowork pairing GPT + Claude for enterprise research. Now it's shipping its own base models. The multi-model enterprise AI stack is taking shape.

🔗 Read the full story → TechCrunch

Stay ahead of AI. Subscribe to LGTM for daily AI model news.