AI Research

Microsoft says MAI beats frontier at 2% of the cost

Microsoft claims MAI Cyber 2 Flash beats GPT-6 Astra at under 2% of task cost, plus steep cuts on coding and image models.

StartupHub.ai Staff

By StartupHub.ai Staff, reviewed by Daniel Singer, Editor

2 min read

Microsoft says MAI beats frontier at 2% of the cost

MAI Cyber 2 Flash beats GPT-6 Astra quality at less than 2% of the cost per task on CyberGym-E2E, Microsoft AI said at Ignite.

That is a pricing claim, not a capability leap.

MAI Code 2 reaches 98% of the coding quality of GPT-6.1 Sol at about 40% of the price, the company said. Against Opus 5.5 it reaches 96% performance for just 5% of the price, and Microsoft positions both Code models for everyday enterprise coding work.

MAI Code 2 Flash matches GPT-6 Luna at 65% of the price and Haiku 5.5 at 20%, Microsoft said. MAI Image 2.7 ranks first on Artificial Analysis image editing and undercuts GPT Image 2.6 Sunburst (max) at one-fifth of the price.

Microsoft AI also listed MAI Realtime 1 from $0.01 per audio minute, Transcribe 2 powering 10 million patient encounters a month, Decision 1.1 at 50x faster and 79x cheaper than GPT-6.1 Sol, and Maia 200 at 1.7x more tokens per watt, the company said.

The numbers do not show evaluation setup, token counts or sustained throughput. Independent cost-per-task accounting would change the read.

Will buyers trade that last 2% to 4% of quality for a 20x cut in agent run cost?

Sources

Get the free daily brief every morning

Every morning the day's AI news, new AI startups and new funding rounds. Free and takes a few minutes to read, unsubscribe at any time.

Join 64,000 other subscribers.

TopicsMicrosoftMAI Cyber 2 FlashMAI Code 2GPT-6 AstraGPT-6.1 SolOpus 5.5MAI Image 2.7Maia 200

StartupHub.ai StaffDaniel SingerWritten by StartupHub.ai Staff, Editorial teamReviewed by Daniel Singer, EditorAbout

StartupHub.ai Staff, Editorial team

The staff writers of StartupHub.ai, ranging from investment analysts to avid AI tool users, early adopters and critical enthusiasts. Backgrounds span engineering, business and the arts. We hold every piece to rigorous standards of research and review.

Daniel Singer, Editor

Daniel Singer is an AI expert who has been writing about, covering and building AI technology since 2018.

More from Daniel Singer

Test your Agent Readiness now

Paste your website. How well ChatGPT, Claude and the other agents can read and use a site, in about a minute.

Watch

The videos behind the news, one flip at a time

The feed mixes the videos from our articles with startup demos and you flip through them like TV channels for free with no signup.