Ignite 2026: MAI Code 2, Office in Copilot
Partly confirmed
Status
- Claim
Partly confirmed
- Our reporting
- Medium confidence
- Importance
- 3 of 5
- Last verified
- 18 November 2026
Your AI and this story
- GPT-6.2 Astra170 days after its cutoff
- Claude Opus 5.5140 days after its cutoff
- Gemini 4 Hydrogen231 days after its cutoff
- Grok 4.7170 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 140 days before it.
Key facts
- Vendor claims (Thurrott, Microsoft live blog): MAI Code 2, 98% of GPT-6.1 Sol’s coding quality at ~40% of the price and 96% of Opus 5.5 at 5% of its price; MAI Code 2 Flash matches GPT-6 Luna at 65% of its price; MAI Cyber 2 Flash (12B active parameters) matches GPT-6 Astra at under 2% of the cost per task, for selected customers first; MAI Realtime 1 from $0.01 per audio minute; Decision 1.1 ‘50x faster and 79x cheaper’ than GPT-6.1 Sol
- MAI Image 2.7: reported top of Artificial Analysis’s image-editing ranking (the only independent number cited)
- All models run in production on Microsoft’s Maia 200 accelerator; ‘MAI Licensed Weights’ for customers to deploy and customise MAI models in their own Azure tenants is ‘coming soon’
- Copilot (Microsoft live blog): new Copilot home (Chat and Cowork) and Office in Copilot generally available; Code and Autopilot rolling out to Frontier customers; Copilot Managed Runtime GA; Microsoft IQ context platform; Agent 365 cost management and analytics; Microsoft Perception security agents GA
- Reuters and GeekWire framed the launch as Microsoft taking on its own partners OpenAI and Anthropic with cheaper, customisable models
- Microsoft AI’s own page counts seven models: MAI Code 2 and Code 2 Flash in GitHub Copilot (Code 2 Flash also in limited preview in Foundry), MAI Image 2.7 in Foundry (Image 2.7 Flash ~$10 per 1,000 images; ‘#1 on Artificial Analysis image editing’), MAI Realtime 1 in MAI Labs, Dragon Copilot Transcribe, Decision 1.1 and MAI Cyber 2 Flash (output tokens 86% cheaper than Cyber 1.1 Flash); vendor benchmarks cited include DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1
What happened
Microsoft used its Ignite keynote to present a full in-house model line-up from Mustafa Suleyman’s Microsoft AI group alongside the OpenAI and Anthropic models it already resells. The pitch is cost: each model is compared with a partner model at a fraction of the price. Coverage counts six or seven models depending on whether the Dragon Copilot transcription product is included. Microsoft’s live blog also announced that Office in Copilot and the new Copilot home are generally available. The Verge reported, from a separate interview, that Satya Nadella now calls himself “AGI-pilled” (headline only; not readable by our fetcher). Before Ignite, Nadella told The Pragmatic Engineer that verification of code becomes more important as generating it gets cheap.
Why it matters
Microsoft is the largest distributor of OpenAI’s models and a major Anthropic customer; shipping cheaper own models for coding and cyber defence puts it in direct price competition with both. No independent coding or cyber benchmark of MAI Code 2 or MAI Cyber 2 Flash was available at launch, so the claims are unverified.
Sources
9 sources from 9 sites. Numbers match the chips in the text.
9 sources: 3 primary, 6 press
Primary
- Microsoft: Ignite 2026 live newsnews.microsoft.com, official
- Microsoft AI: Ignite 2026, seven new modelsmicrosoft.ai, official
- Microsoft Security blog: Security for the agentic era, powered by Microsoft Perceptionmicrosoft.com, official
Press
- Thurrott: Ignite 2026: Microsoft AI announces 7 new modelsthurrott.com, press
- Axios: Microsoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsaxios.com, press
- Korea Times (Reuters): Microsoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIkoreatimes.co.kr, press
- Ars Technica: Copilot is ‘the new OS for work,’ and now Office runs inside of itarstechnica.com, press
- The Pragmatic Engineer: Q&A with Satya Nadella on software engineering after code becomes cheapnewsletter.pragmaticengineer.com, press
- GeekWire: Microsoft jabs at OpenAI and Anthropic with cheaper modelsgeekwire.com, press
Changes
- Microsoft AI’s model page (availability, Image 2.7 Flash price)
- Filed from Microsoft’s live blog and press (Axios returned 403; Ars and The Verge not readable, cited by headline)