Microsoft Unveils Specialized AI Models at Ignite
At Microsoft’s Ignite conference, the company introduced a new lineup of purpose‑built AI models aimed directly at Anthropic and OpenAI in coding, cybersecurity, and image generation. Rather than chasing a single monolithic chatbot, Microsoft is betting on smaller, task‑focused models that deliver comparable performance at a substantially lower cost.
MAI Code 2 Sets New Benchmarks for Developers
MAI Code 2 is positioned as a cost‑effective alternative to leading coding assistants. Microsoft claims it achieves 96% of Claude Opus 5.5’s performance on common enterprise coding tasks while costing only 5% of the price. The model also matches 98% of GPT‑6.1 Sol’s output at 40% of the price. For routine fixes, the lighter MAI Code 2 Flash model targets everyday edits, offering performance comparable to Claude Haiku 5.5 at just 20% of the cost. Both coding models are already integrated into GitHub Copilot, bringing Microsoft’s in‑house technology directly into a product that has traditionally leaned on third‑party AI providers.
OpenAI also released GPT‑6.2 on Tuesday, so Microsoft’s Sol comparison already concerns an older generation. The figures are based on four evaluations covering practical software development tasks such as fixing code and completing multi‑file changes. Microsoft ran three of the tests on its own platform, used external results for the fourth, and applied a 50 % promotional discount when calculating MAI Code 2’s costs.
MAI Cyber 2 Flash Closes the Gap on Vulnerability Detection
Microsoft’s security offering, MAI Cyber 2 Flash, claims a 92.3 % success rate on the CyberGym‑E2E vulnerability‑discovery benchmark, edging out GPT‑6 Astra’s 91.5 % while costing less than 2 % of the per‑task expense. CyberGym‑E2E measures the percentage of security investigations an AI completes successfully; Microsoft excluded 92 of 905 tasks after identifying overlap with its training data. The model will begin replacing an older MAI security system for eligible MDASH customers in mid‑December.
MAI Image 2.7 Leads in Independent Image‑Editing Rankings
The most compelling evidence for Microsoft’s specialization strategy may lie in its imaging model, MAI Image 2.7. Quick side‑by‑side tests show it outperforming GPT Image and Google’s Nano Banana on realism, creativity, and overall quality. The model currently tops Artificial Analysis’s image‑editing leaderboard with a 1,215 Elo score, narrowly ahead of GPT Image 2.6 Sunburst’s 1,211. While the text‑to‑image generation ranking places Microsoft fourth, the narrow margin suggests the competition is tight.
Pricing underscores the value proposition. Microsoft quotes roughly $35 per 1,000 images for MAI Image 2.7 and $10 for its Flash version, compared with about $211 for GPT Image 2.6 Sunburst at maximum quality on Artificial Analysis’s benchmark. Google’s Nano Banana 2.1 sits near $34 per 1,000, while its Pro variant is around $134.
Microsoft is also offering select business customers the ability to customize and run licensed versions of its models inside Azure, giving enterprises more control while sidestepping reliance on external vendors. Code 2 is already available in GitHub Copilot, and the Cyber 2 Flash rollout is set to begin in mid‑December for MDASH customers.
Also Read
- Sim-to-Real Transfer Gives AI Reasoning Model Real-World Robot Skills
- Williams Signals Gradual Fed Tightening, Pushes Back on Aggressive December Hike
- Bitcoin Policy Institute Rebrands as Finney Foundation, Expands Mission to Open-Source Technology
- Crypto Roundup: Solana’s CoinGecko Stake, Coinbase’s New York Lending Rollout, and Fresh ETH Transfers Tied to Binance Theft


