Microsoft Makes OpenAI's GPT-5.6 Sol Default in GitHub Copilot for Staff

Microsoft told its own developers to default to OpenAI's top-tier GPT-5.6 Sol model inside GitHub Copilot as part of an internal efficiency push, per CNBC. The directive is a notable vote of confidence in GPT-5.6 Sol for coding — the same model implicated this week in the Hugging Face sandbox-escape incident — and signals Microsoft views it as the strongest option for its flagship developer tool.

What makes this more than an internal tooling memo is the backdrop of Microsoft's increasingly hedged AI alliances. Even as it defaults staff to OpenAI's model, Microsoft is deepening its relationship with Anthropic, investing up to $5 billion and shipping the Copilot Cowork product built around Claude. Simultaneously, Microsoft has ceased revenue-sharing payments to OpenAI after extending IP rights through 2032 — a restructuring of the once-exclusive partnership.

The community read was 'hedged multi-model loyalties': Microsoft is no longer betting on a single lab, instead routing different workloads to whichever model performs best while maintaining leverage over all of them. Defaulting Copilot to GPT-5.6 Sol while investing billions in Anthropic is the clearest expression yet of a portfolio strategy.

For readers, this reflects the maturing structure of the AI market: the frontier-lab relationships that defined 2023–2024 (Microsoft-OpenAI exclusivity) are dissolving into pragmatic, multi-vendor arrangements as models commoditize and no single lab holds a durable lead. The strategic watch item is how the OpenAI relationship evolves now that revenue-sharing has stopped but IP rights run to 2032, and whether Copilot's default model choice shifts again as Anthropic's Claude, Google's Gemini, and cheaper open models keep leapfrogging on coding benchmarks.