Choose Rich Live
Tech

Microsoft debuts in-house MAI models targeting rival performance at lower costs

At its Ignite conference, Microsoft introduced a new suite of in-house MAI artificial intelligence models spanning coding, cybersecurity, image editing, and speech, claiming significant cost efficiencies over rivals OpenAI and Anthropic.

Arena benchmark chart showing MAI-Image-2.7 ranked second with a score of 1,475 on the Image Edit Arena leaderboard.
Image: @arena

Microsoft announced a lineup of first-party MAI artificial intelligence models during its Ignite conference, directly challenging offerings from Anthropic and partner OpenAI. The suite spans programming, cybersecurity, voice, and image generation, released on the same day OpenAI rolled out GPT-6.2.[2][7][9]

Leading the release is MAI Code 2, integrated into GitHub Copilot. Microsoft claims the model delivers 98% of OpenAI's GPT-6.1 Sol performance at 40% of the cost, while reports cite company comparisons showing it rivals Anthropic's Claude Opus 5.5 at just 5% of the expense. Microsoft also introduced MAI Code 2 Flash, which AT&T testing showed outperforming GPT-6.1 Sol at one-fiftieth the cost, and MAI Cyber 2 Flash, which surpassed GPT-6 Astra on security benchmarks at under 2% of the cost per task.[3][4][5][6][7][8]

The company also rolled out MAI Realtime 1 and slated MAI Transcribe 2 for Dragon Copilot's medical workflows. Meanwhile, evaluation platform Arena reported that MAI Image 2.7 has exited preview, placing Microsoft second across all three Image Arenas. Priced at $0.035 per image, the model scored 1,475 on Image Edit Arena, trailing only GPT Image 2.5 Sunburst at 1,519 and topping Grok Imagine Image 2.0 at 1,437.[1][7]

Key facts

  • Microsoft announced six new in-house MAI models covering coding, cybersecurity, image generation, and audio at its Ignite event.
  • MAI Code 2 is available in GitHub Copilot, with Microsoft claiming 98% of GPT-6.1 Sol's performance at 40% of the cost, and rivaling Claude Opus 5.5 at 5% of the cost.
  • MAI Code 2 Flash outperformed GPT-6.1 Sol in AT&T evaluations while costing 50 times less, according to Microsoft.
  • MAI Cyber 2 Flash outperformed GPT-6 Astra on a cybersecurity benchmark at under 2% of the cost per task.
  • MAI Image 2.7 costs $0.035 per image and ranks second on Arena's Image Edit leaderboard with a 1,475 score, behind GPT Image 2.5 Sunburst (1,519).
  • Microsoft plans to deploy MAI Transcribe 2 to Dragon Copilot for clinical and medical use cases.
  • The MAI model announcements arrived on the same day partner OpenAI launched GPT-6.2.

Sources · 9 sources

  1. AA

    Arena.ai@arenaPost on X ·

    MAI-Image-2.7 from @MicrosoftAI is out of preview, making Microsoft AI the #2 lab across all 3 Image Arenas. It’s also on the Pareto frontier across all at $0.035 per image, providing a strong cost-effective option. The latest model lands #3 on Multi-Image Edit, #4 on Single Image Edit, and #4 on Text-to-Image. That’s roughly ~+50 pts over MAI-Image-2.6 across Image Arenas. Its largest gains vs 2.6 are in Single Image Edit performance: - #3 Photorealistic & Cinematic Imagery (from #8) - #4 Text Rendering (from #13) - #4 Product, Branding & Commercial Design (from #10) - #4 3D Imaging & Modeling (from #9) - #4 Portraits (from #9) - #5 Cartoon, Anime & Fantasy (from #11)

    Open source
  2. RE

    Reuters@ReutersPost on X ·

    Microsoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAI https://t.co/f1D1rOWIbj

    Open source
  3. SB

    Shay Boloor@StockSavvyShayPost on X ·

    $MSFT just unveiled six in-house AI models at Ignite including MAI Code 2 which delivers 98% of GPT-6.1 Sol’s performance at 40% of the cost. And with that model already in GitHub Copilot Microsoft can shift more inference onto its own models and keep more of the margin. https://t.co/GHQyVFkfxi

    Open source
  4. LD

    Lark Davis@LarkDavisPost on X ·

    Microsoft just launched a suite of new AI models at Ignite 🚀 - MAI Code 2: coding model heading to GitHub Copilot - MAI Code 2 Flash: cheaper, faster version - MAI Cyber 2 Flash: security model, limited customers only - New image, voice and transcription models Microsoft says Code 2 hits 96% of Claude Opus 5.5’s performance at 5% of the cost.

    Open source
  5. ZD

    ZDNET@ZDNETPost on X ·

    Microsoft says its new AI coding model rivals Claude Opus 5.5 at 5% of the cost - ZDNET https://t.co/OE1W4mQg7E

    Open source
  6. TE

    Techmeme@TechmemePost on X ·

    Microsoft unveils six AI models spanning coding, cybersecurity, images, and voice, saying MAI Code 2 has 98% of GPT-6.1 Sol's performance at 40% of the price (@inafried / Axios) (Visit Techmeme dot com for the link and full context!)

    Open source
  7. WS

    Wall St Engine@wallstenginePost on X ·

    MICROSOFT ANNOUNCES 6 NEW AI MODELS AT IGNITE $MSFT unveiled six MAI models across coding, cybersecurity, image generation and voice on the same day its partner OpenAI released GPT-6.2. 1) MAI Code 2: Delivers 98% of GPT-6.1 Sol's performance at 40% of the cost. Now in GitHub Copilot. 2) MAI Code 2 Flash: Outperformed GPT-6.1 Sol in AT&T testing while costing 50 times less. 3) MAI Cyber 2 Flash: Outperformed GPT-6 Astra on a cybersecurity benchmark at less than 2% of the cost per task. 4) MAI Realtime 1: Outperformed GPT Realtime 2.1 on public benchmarks and Microsoft's own evaluations. Microsoft is also launching MAI Image 2.7 in public preview and bringing MAI Transcribe 2 to Dragon Copilot for medical use cases. Separately, Microsoft says it is making progress on its own general-purpose frontier AI model.

    Open source
  8. CB

    Crypto BriefingArticle ·

    Microsoft launches MAI Code 2 and MAI Code 2 Flash on GitHub Copilot Microsoft's new models could reshape AI coding economics, offering enterprises cost-effective alternatives, but independent validation is pending. The post Microsoft launches MAI Code 2 and MAI Code 2 Flash on GitHub Copilot appeared first on Crypto Briefing .

    Open source
  9. TE

    TechmemeArticle ·

    Microsoft unveils six AI models spanning coding, cybersecurity, images, and voice, saying MAI Code 2 has 98% of GPT-6.1 Sol's performance at 40% of the price (Ina Fried/Axios) Ina Fried / Axios : Microsoft unveils six AI models spanning coding, cybersecurity, images, and voice, saying MAI Code 2 has 98% of GPT-6.1 Sol's performance at 40% of the price — Microsoft AI CEO Mustafa Suleyman tells Axios the software giant is making significant progress toward building …

    Open source