Skip to content
Rush Commerce
Tools & Teardowns3 min read

Microsoft MAI Code 2 and Realtime 1: read the price footnotes

At Ignite 2026 Microsoft launched MAI Code 2, MAI Realtime 1 at $0.01/min plus reasoning tokens, and MAI Image 2.7. The cheap prices come with fine print.

Microsoft opened Ignite 2026 with a batch of new MAI models aimed at one pitch: near-frontier quality for a fraction of the price. The Microsoft AI announcement covers MAI Code 2, MAI Code 2 Flash, MAI Realtime 1, MAI Image 2.7 and MAI Cyber 2 Flash. The prices are real. So are the footnotes, and that is where a small business should spend its five minutes.

What actually happened

From Microsoft's own posts, all dated November 17:

  • MAI Code 2 is a coding model. Microsoft says it delivers 98% of GPT-6.1 Sol's performance at about 40% of the cost, averaged over four coding benchmarks. The footnote: that cost uses Code 2's 50% promotional price. No per-token list price is published. It is available in GitHub Copilot; MAI Code 2 Flash is also in limited-access preview in Microsoft Foundry.
  • MAI Realtime 1 is Microsoft's first speech-to-speech model for live voice agents. The headline is $0.01 per minute. The footnote adds reasoning billed separately at $1 per million input tokens and $5 per million output. It is in limited-access preview in Foundry through an interest form, and it is compatible with the OpenAI Realtime API. Preview users are not billed until public preview.
  • MAI Image 2.7 runs about $35 per 1,000 images and 2.7 Flash about $10 per 1,000, per Microsoft. Both are in public preview on Foundry and through fal, OpenRouter, Vercel and Genspark.
  • MAI Cyber 2 Flash lists at $0.10/$0.50 per million tokens but ships only inside Microsoft security products, starting mid-December.

Every quality number above is Microsoft's own benchmark. We found no independent tests yet.

Why it matters for your business

Two of these are worth a look for a small operator.

Voice agents. A phone or booking agent at a cent a minute is a different business case than one at 10x that. But "starts at $0.01" is not your bill. A voice agent that checks inventory or looks up an order will use reasoning tokens on most calls. Because Realtime 1 speaks the OpenAI Realtime API, you can test it against your current provider without a rewrite, which is the right way to find the real per-call cost.

Coding. A 50% promotional price is a temporary price. If you build cost estimates on it, write the expiry risk into the estimate.

Our rule: price every model by cost per finished task, at list price, with the reasoning tokens in. Then decide.

Key takeaways

  • MAI Code 2's "40% of the cost" claim uses a 50% promotional price; no list price is published
  • MAI Realtime 1 is $0.01/min plus $1/$5 per million reasoning tokens, in limited preview
  • Realtime 1 is OpenAI Realtime API-compatible, so A/B testing is cheap
  • MAI Image 2.7 Flash runs about $10 per 1,000 images in public preview
  • All benchmarks are Microsoft's; measure cost per finished task yourself

Want a voice agent priced on real calls, not a launch slide? We build voice and support agents on swappable providers and measure cost per resolved call before you commit. Tell us what your phones handle or run the numbers.

Sources: Microsoft AI — Ignite 2026, Microsoft AI — MAI Code 2, Microsoft AI — MAI Realtime 1, Microsoft AI — MAI Image 2.7.

  • #microsoft
  • #mai-code-2
  • #mai-realtime-1
  • #ignite-2026
  • #ai-costs
TR

Tommy Rush — Founder, Rush Commerce

Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More

Get The Rush Report weekly — one email, zero fluff.