Microsoft releases seven MAI models at Ignite, led by MAI Code 2, and opens licensed weights
Microsoft has released a family of seven new models from its Microsoft AI (MAI) group at Ignite, aimed at matching rivals on everyday tasks at a much lower price, according to the company's announcement. It also announced MAI Licensed Weights, which will let companies customise MAI models with their own data and run them in their own Azure environments.
The centrepiece is MAI Code 2, which Microsoft says reaches 98% of the coding performance of OpenAI's GPT-6.1 Sol at about 40% of the price and 96% of Anthropic's Claude Opus 5.5 at 5% of the price, averaged over four coding benchmarks. Microsoft says it is tuned for everyday enterprise coding and explicitly not yet for harder work such as science research or very long agentic tasks. It is available in GitHub Copilot today, along with a smaller MAI Code 2 Flash.
The other releases are MAI Cyber 2 Flash, which Microsoft says matches GPT-6 Astra on the CyberGymE2E security benchmark at under 2% of the cost per task; MAI Image 2.7, which it says ranks first on Artificial Analysis for image editing; MAI Realtime 1, its first live speech-to-speech model, priced from $0.01 per audio minute; Dragon Copilot Transcribe, built on MAI Transcribe 2; and Decision 1.1, a model that routes requests to bigger models only when needed. Microsoft also says its second-generation Maia 200 AI chip is now in production running its largest models.
MAI Licensed Weights is coming soon through Microsoft's Frontier Company engagements. Microsoft says customers' data and adaptations stay theirs and are used to train its models only with their express agreement, and it cites customer results such as AT&T, which it says beat GPT-6.1 Sol on standard 5G network analytics tasks at a fiftieth of the cost, according to its Licensed Weights post.
"This isn't just one model. It's not a flash in the pan," Microsoft AI chief executive Mustafa Suleyman said. BigGo Finance notes that the performance and pricing figures are Microsoft's own claims on selected benchmarks, and that some price advantages reflect promotional discounts, according to BigGo Finance.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Cybersecurity AI Weekly, Silicon Brief or Daily AI Brief.
Also, consider following us on social media:
More from Cybersecurity
Nov 17 Microsoft launches AI Transparency Initiative with more than 40 organisations Nov 17 Gemini AI agents in safety tests repeatedly bypassed safeguards to attack Epoch AI's live website Nov 17 Anthropic cuts its vulnerability disclosure deadline to 30 days for serious bugs Nov 17 OpenAI gives NATO access to its Daybreak cyber defense models Nov 17 Anthropic flagged its AI agents' fake records requests to Virginia's education departmentMore from Data Centers
Nov 17 HSBC sells $300 million of Oracle-linked Project Jupiter data centre loan at a discount Nov 17 Siemens to acquire Breker Verification Systems to expand chip verification for AI designs Nov 17 Argentum AI signs $630 million, 20-year data centre deal with Kyndryl for Nvidia B300 systems Nov 17 Ramon.Space raises $124 million for computers that run AI in orbit Nov 17 Q.ANT unveils NPU Gen 2+ photonic processor and runs an LLM with optical Fourier transformsAI Programming Weekly
Weekly news about AI tools for software engineers, AI enabled IDE's and much more.
Whitepaper
Tensordyne Napier: What If One Rack Could Do the Work of Nine?
Tensordyne
This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.
Read moreYou may also like
Google Cloud Adds OpenAI Models to Model Garden
Artificial Analysis Index v5: open-weight models fall 12 points behind
Alibaba launches Qwen4-Max, a closed 2.4-trillion-parameter flagship model
Microsoft launches AI Transparency Initiative with more than 40 organisations
Liquid AI Releases LFM3, Six Open-Weight Models for On-Device AI
Daily AI Brief: the AI news that matters, in your inbox.