The number I care about in Microsoft's Nov.
The number I care about in Microsoft's Nov. 17 MAI Code 2 announcement is the price. Microsoft is coming for the everyday coding work that pays the AI labs' bills, and it already sells those labs' models inside its own products.
- Microsoft says MAI Code 2 delivers 96% of Claude Opus 5.5's performance at 5% of the cost, and 98% of GPT 6.1 Sol's at 40%, averaged across four coding benchmarks. Those costs use a 50% promotional price, and Microsoft says results vary by workload and configuration.
- Both models are available in GitHub Copilot now. Flash is in limited-access preview in Foundry, with OpenRouter, Vercel and others described as coming. Microsoft is also deprecating MAI-Code-1.1-Flash in December.
- OpenAI announced GPT-6.2 Astra and Sol the same day, and Microsoft benchmarked against 6.1. A third-party head-to-head with 6.2 Sol and Opus 5.5 is what I'm waiting for.
- Microsoft says blind reviewers preferred MAI Code 2 over GPT-6 Sol 67% of the time. It also says neither model is tuned for science research or very long-horizon agent work.
$MSFT closed at $506.24, down 2.27%, and traded at $507 after hours. Nothing ties that move to this announcement. My read: once the promo pricing ends, even a cheaper-but-close model in Copilot pressures lab pricing into any IPO talk. I could be wrong if independent tests close the gap.
Source:
Microsoft AI: Introducing MAI Code 2 and Code 2 Flash: https://microsoft.ai/news/introducing-mai-code-2-and-mai-code-2-flash/
Not financial advice. For informational and educational purposes only.