Home » Microsoft Unveils Seven MAI Models at Ignite to Undercut OpenAI and Anthropic on Price

Microsoft Unveils Seven MAI Models at Ignite to Undercut OpenAI and Anthropic on Price

By

–

Microsoft Unveils Seven MAI Models at Ignite to Undercut OpenAI and Anthropic on Price

SEATTLE, November 19, 2026 — Microsoft AI unveiled seven new MAI models at its Ignite conference, promising frontier‑level performance at a fraction of the price charged by OpenAI and Anthropic. The headline model, MAI Code 2, is billed to deliver 96% of Claude Opus 5.5’s coding capability while costing just 5% of the rival’s price.

Price‑Performance Push Across Six Domains

Microsoft’s rollout spans coding, cyber defense, model routing, image generation, live speech and medical transcription. The company says the lineup is built “around price‑performance” to let enterprises run agents 24‑7 on real engineering tasks. The new coding models—MAI Code 2 and MAI Code 2 Flash—target everyday developer workloads, from deep debugging to high‑volume code explanations.

According to Microsoft, MAI Code 2 hits 96% of Claude Opus 5.5’s performance on four coding benchmarks (DeepSWE 1.1, Terminal‑Bench 2.1, SWE‑Atlas and FrontierCode v1.1) while costing about 5% of the price. In blind frontend tests, developers chose MAI Code 2 over OpenAI’s GPT‑6 Sol 64% of the time. The Flash variant matches Claude Haiku 5.5 at roughly 20% of the cost.

Strategic Shift From External Partners

The launch marks another step in Microsoft’s effort to reduce reliance on external AI providers. Analyst Pareekh Jain notes the move “protects its profits and locks developers into its own tools,” especially given Microsoft’s deep integration with GitHub, Visual Studio and Azure. By offering in‑house models, Microsoft aims to avoid paying OpenAI and Anthropic for every GitHub Copilot request.

Microsoft also introduced a Licensed Weights program that lets enterprises customize frontier MAI models with proprietary data and run them securely in Azure. “You control who can access it. Your data and adaptations stay entirely yours,” said Mustafa Suleyman, CEO of MAI. Early adopters such as AT&T reportedly built a custom MAI model for 5G analytics that beat GPT‑6.1 Sol at 50× lower cost.

Cost Controls Built Into the Stack

Beyond cheaper models, Microsoft added a routing layer—Decision 1.1—that forwards tasks to larger, more expensive models only when needed, according to the company’s AA Decision Index ranking. The new cyber‑defense model, MAI Cyber 2 Flash, claims token output that is 86% cheaper than its predecessor, MAI Cyber 1.1 Flash.

All benchmark and pricing claims are Microsoft’s own and have not been independently verified.

Industry Context: Rival Moves and IPOs

The same week OpenAI released GPT‑6.2 Astra and Sol, while Anthropic priced its Nasdaq debut at $83 per share, aiming for a post‑IPO valuation of about $1.5 trillion. Microsoft’s aggressive pricing appears designed to undercut those rivals on enterprise workloads, especially as AI‑driven agent usage expands.

What’s Changing for Developers

That shift could make AI‑assisted development more accessible to midsize firms that previously found frontier models prohibitively expensive.

Microsoft says the next phase of its MAI roadmap includes broader availability of Licensed Weights and further optimization of its second‑generation Maia 200 AI accelerator, now in production with 1.7× more tokens per watt.

Sources

Photo: BCS (PD)

Get breaking stories first on our Telegram channel.

More stories by

Most read

AI Tools

Artificial Intelligence

Business

World News