Microsoft has rolled out MAI-Code-1-Flash, its first in-house coding model, to every GitHub Copilot plan as part of Microsoft Build 2026. The model is built end-to-end by Microsoft on clean, appropriately licensed data, continuing the company's effort to wean its flagship developer products off third-party model dependencies.
The performance numbers tell a tight story. MAI-Code-1-Flash outperforms Claude Haiku 4.5 across all four core coding benchmarks Microsoft tested, with the biggest lead — a 16-point margin — on SWE-Bench Pro, the toughest of the four, where it scores 51.2 percent versus Haiku 4.5's 35.2 percent. Just as important for production use, Microsoft says the model solves harder problems with up to 60 percent fewer tokens than comparable small coding models, which translates directly into latency and cost wins for high-volume developer workflows.
For GitHub Copilot users, the rollout is immediate and universal. Every paid plan now has access to MAI-Code-1-Flash as an option, sitting alongside existing OpenAI and Anthropic models in the model picker. Microsoft is positioning Flash for everyday inline completions, edits, and chat — the high-frequency, latency-sensitive parts of the Copilot loop — while heavier reasoning workloads route to MAI-Thinking-1 or third-party frontier models.
The release marks the start of a broader Microsoft AI Model family. The company is signaling that it now considers in-house frontier model development a strategic capability, not a hedge, and that GitHub Copilot — its single largest AI developer surface — will be the proving ground.
Source: [Microsoft AI](https://microsoft.ai/news/introducingmai-code-1-flash/)