The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard
Minor World

Microsoft unveils seven MAI models at Ignite, led by cheaper coding model Code 2

By The Daily Byte ·20 Nov 2026 ·5 min read
GitHub Copilot chat panel open beside JavaScript code in Visual Studio Code
Microsoft has unveiled seven new artificial intelligence models built by its own researchers at its Ignite 2026 conference in San Francisco, in the company's biggest push yet to compete with OpenAI and Anthropic on price. The launch, announced on 17 November, covers coding, cybersecurity, image generation, live voice, transcription and decision-making. The headline model is MAI Code 2, a coding model Microsoft says reaches 98% of OpenAI's GPT-6.1 Sol on coding tests at roughly 40% of the price, and 96% of Anthropic's Claude Opus 5.5 at about 5% of the price. Those figures come from Microsoft's own testing and have not been independently verified. ## Seven models, six kinds of work The seven models announced at Ignite are: - **MAI Code 2**, for everyday enterprise software work, including debugging and work across multiple files. - **MAI Code 2 Flash**, a lighter model for routine jobs such as code explanations and quick fixes. - **MAI Cyber 2 Flash**, a security model built to scan large codebases around the clock for weaknesses. - **MAI Image 2.7**, for image generation and editing, with a cheaper Flash variant alongside it. - **MAI Realtime 1**, Microsoft's first live speech-to-speech voice model. - **MAI Transcribe 2**, which powers Dragon Copilot Transcribe for clinical documentation. - **Microsoft-Decision-1.1**, a small routing model that decides which model should handle each task. Microsoft also put its Maia 200 chip, its second-generation AI accelerator, into production, and introduced MAI Licensed Weights, a programme that lets selected enterprises tune an MAI model on their own data and run it inside their own Azure environment. ## MAI Code 2: the pitch is the bill MAI Code 2 is the model developers are most likely to meet first. It is live now in GitHub Copilot, Microsoft's AI coding assistant, across paid Copilot plans. The lighter Flash version is available from the Free tier upwards, and in a limited-access preview on Microsoft Foundry, the company's platform for building AI applications. Microsoft says Code 2 averaged 98% of GPT-6.1 Sol's score across four coding benchmarks — DeepSWE, Terminal Bench, SWE Atlas and FrontierCode — at about 40% of the price, and 96% of Claude Opus 5.5 at about 5% of the price using a promotional rate. In a separate human test, reviewers preferred Code 2's output over GPT-6.1 Sol's 64% of the time, across 258 annotators. The promotional pricing is time-limited. Code 2 costs $0.50 per million input tokens and $2.50 per million output tokens until 17 May 2027, after which the prices double to $1.00 and $5.00. Flash costs $0.05 and $0.25 per million tokens on the same promotional terms. Microsoft itself notes that Code 2 is not yet tuned for science research or very long-running agent tasks. No benchmark figure in the announcement has been checked by an independent laboratory. ## Security, images and voice The security model, MAI Cyber 2 Flash, is aimed squarely at defensive work. Microsoft says it beats GPT-6 Astra on the CyberGym end-to-end security benchmark at less than 2% of the cost per task, again on the company's own measurements. It has a one-million-token context window and is listed at $0.10 per million input tokens and $0.50 per million output tokens. It runs only inside Microsoft security products such as MDASH, with the new model replacing the current one in eligible profiles from mid-December. MAI Image 2.7 is ranked first on Artificial Analysis's image-editing leaderboard, according to Microsoft, ahead of GPT Image 2.6 at about a fifth of the price. It is in public preview on Foundry at about $35 per 1,000 images, with the Flash variant at about $10. MAI Realtime 1, the company's first speech-to-speech model, starts at $0.01 per audio minute and is available in MAI Labs and a limited Foundry preview. Dragon Copilot Transcribe, built on MAI Transcribe 2, already handles ten million patient encounters a month, Microsoft says. Decision 1.1, the routing model, is generally available now. Microsoft says it is 50 times faster and 79 times cheaper than GPT-6.1 Sol for the small decisions — routing, classification and verification — that keep agent workflows running. ## What it means The strategy is plain: Microsoft is betting that most business AI work does not need the most powerful model available, and that customers will choose a model that is nearly as good at a fraction of the price. By building the models in-house under its Microsoft AI division, led by chief executive Mustafa Suleyman, the company also reduces its dependence on OpenAI, in which it remains a major investor and partner. Whether the price advantage holds up in real use will depend on independent testing that has not yet happened. For now, developers using GitHub Copilot can try Code 2 on their own code and judge for themselves. [Source: Orca Router](https://www.orcarouter.ai/blog/mai-code-2-ignite-seven-models) and [The Latent Daily](https://thelatentdaily.com/news/microsoft-mai-models-ignite-2026-licensed-weights/), reporting Microsoft's Ignite 2026 announcement of 17 November 2026. ## Frequently asked questions **What is MAI Code 2?** MAI Code 2 is Microsoft's new in-house coding model, launched at Ignite 2026. It is designed for everyday enterprise software work, including debugging and multi-file coding, and is available in GitHub Copilot. **How many MAI models did Microsoft launch at Ignite 2026?** Seven: MAI Code 2, MAI Code 2 Flash, MAI Cyber 2 Flash, MAI Image 2.7, MAI Realtime 1, MAI Transcribe 2 and Microsoft-Decision-1.1. **Is MAI Code 2 free?** The lighter MAI Code 2 Flash is available to GitHub Copilot Free users. The full MAI Code 2 model is available on paid Copilot plans.