Skip to content
NewA calmer, faster ZeroHour

Microsoft unveils seven in-house models to replace partner AI

Original title: Microsoft is replacing outside AI models in its own products with cheaper in-house alternatives

The Decoderby Jonathan Kemper ()7 min read
Part of a story covered by 3 sourcesMicrosoft unveils seven cheaper in-house MAI models at IgniteMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

Microsoft says MAI Code 2 averages 98 percent of GPT-6.1 Sol at 40 percent of the cost.

At Ignite, Microsoft AI released seven models for code, security, images, speech, and routing, aimed at always-on agents and priced against OpenAI and Anthropic models. The company says MAI Code 2 averages 98 percent of GPT-6.1 Sol across four benchmarks at 40 percent of the cost, using a 50 percent promotional discount. MAI Cyber 2 Flash and a Gemma-based decision model are also priced far below compared models. Agentic tool benchmarks are still unpublished.

Unconfirmed: Benchmark, cost, and preference figures are Microsoft's; a GPT-6 Cyber comparison and agentic results are not yet published.

Key facts
  • Promotional MAI Code 2 pricing runs through May 17, 2027.
  • Microsoft says MAI Cyber 2 Flash solved 92.3 percent of CyberGym-E2E tasks.
  • About 10 percent of CyberGym tasks overlapped training sources and were removed.
  • Microsoft-Decision-1.1 is based on Google's Gemma 4 26B A4B.
  • Decision model output tokens are free, Microsoft says.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

7 min read at the-decoder.com

Microsoft's AI division MAI unveiled seven new models for code, security, images, language, and routing at Ignite. MAI Code 2 reaches 98 percent of GPT-6. 1 Sol's performance at 40 percent of the cost, according to Microsoft.

The Decoder

By Jonathan Kemper · the-decoder.com

Read the full article at The Decoder

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.