Microsoft says its own coding model delivers 96% of Opus 5.5 at 5% of the price as it leans less on OpenAI and Anthropic
First seen on X 44 hours ago@satyanadella ♥ 3387 views
Microsoft unveiled an in-house coding model it says comes close to Anthropic's Claude Opus 5.5 at a twentieth of the price. A company that has been paying OpenAI and Anthropic for tokens is now competing with them directly.
At Ignite 2026 in San Francisco on Nov. 17, Microsoft introduced MAI Code 2, Digital Times and ZDNET reported. Microsoft says that across four benchmarks of everyday enterprise development, the model delivers 96% of Opus 5.5's performance at 5% of the price and 98% of OpenAI's GPT-6.1 Sol at 40%. It is available in GitHub Copilot now. Its security model, MAI Cyber 2 Flash, matched GPT-6 Astra's quality on a security benchmark at under 2% of the per-task cost, Microsoft said. It also launched MAI Licensed Weights, letting enterprises fine-tune MAI models on their own data and run them in Azure.
New work tools came alongside: Copilot Workspace, where several employees and their agents share context, and Copilotforce, which brings Salesforce customer data into Copilot. "Until now it's been mostly a single-player world of a person and their agent," CEO Satya Nadella said. "Now we need a true multiplayer mode." Nadella opened his keynote by acknowledging last week's outage that took down Outlook and other services in some 70 countries, saying the company knows it has work to do.
The numbers are Microsoft's own. ZDNET noted that "the benchmarks are all Microsoft-provided" and may differ in independent tests, and that this does not mean a $100-a-month subscription will now cost $5. Still, routing the flood of routine requests through its own models could sharply cut what Microsoft pays outside model makers. Microsoft is also bringing GPT-6.2, released the same day, and Meta's newest models into its Foundry platform.
Comments 0
Select a sentence in the story (long-press on phones) to comment on it.
“With Meta’s latest models coming to Foundry, we’re giving developers and enterprises even more choice and flexibility!”
