Oossa

Microsoft rolls out seven cheaper in‑house AI models at Ignite

Microsoft replaces many third‑party AI services with its own models, promising 40‑98% of the performance at a fraction of the cost.

By Published by Oossa: 2 min read

Surface · Unsplash

At the Ignite conference on Nov 17, 2026 Microsoft announced a suite of seven new AI models. The company says the models are built to cut token costs for always‑on agents while keeping most of the quality of leading competitors like OpenAI’s GPT‑6 series and Anthropic’s Claude. All seven models were released together, rather than over the year as Microsoft did at its Build event in June.

What the new models do

MAI Code 2 targets everyday coding tasks. Microsoft reports it reaches 98% of GPT‑6.1 Sol’s benchmark scores while costing about 40% of the price, and 96% of Claude Opus 5.5’s performance for only 5% of the price. The Flash variant, aimed at explanations and small edits, matches GPT‑6 Luna at 65% of the cost.

MAI Cyber 2 Flash is a security model. On the CyberGym‑E2E benchmark it solves 92.3% of tasks, a slight edge over GPT‑6 Astra, but at less than 2% of the per‑task cost. Token fees are a hundredth of GPT‑6 Astra’s.

Decision 1.1, built on Google’s Gemma 4 26B, handles routing and classification. Microsoft says it is 2.2 × faster and 4.4 × cheaper than comparable large language models, and scores 68 points on Artificial Analysis’s Decision Index, six points higher than its predecessor.

MAI Image 2.7 leads the Artificial Analysis image‑editing leaderboard, beating GPT‑Image 2.6 Sunburst (max) while costing one‑fifth of the price. A cheaper Flash version powers the Canvas editor.

MAI Realtime 1 is a speech‑to‑speech model compatible with OpenAI’s Realtime API. Microsoft claims it outperforms GPT‑Realtime 2.1 and starts at $0.01 per audio minute.

How and where to get them

The models are available now in various Microsoft products. Code 2 and its Flash variant appear in GitHub Copilot, with Flash also in a limited Foundry preview. Decision 1.1 is generally available through Foundry, OpenRouter and Vercel’s AI Gateway. Image 2.7 and Flash are in public preview on Foundry and on third‑party platforms such as Vercel. Realtime 1 can be tried for free in MAI Labs and will move to public preview later. Cyber 2 Flash is bundled with Microsoft’s MDASH security service and will replace the existing model for eligible customers in mid‑December.

Why it matters

Enterprises can keep AI agents running 24/7 without the token bills that have risen sharply. Developers get cheaper code‑generation and security tools that perform almost as well as premium services, which could lower cloud costs for many businesses.

Was this article useful?
Share

Read next

Oossa · Newsletter

The week in AI, explained

Every Monday: the stories worth knowing, in plain language. Free, no spam.