OpenAI: GPT-6 Sol Pro

GPT-6 Sol Pro sits inside OpenAI's three-level GPT-6 family as the "pro" variant of the Sol tier, a step designed for strong reasoning on demanding tasks while remaining a single shared endpoint with its siblings. The pro reasoning mode is selected either by setting reasoning.mode to "pro" on the standard Sol identifier or by switching to the Sol Pro slug directly, mirroring the same toggle pattern used in the prior generation. Because all three GPT-6 tiers share one API, applications can move between Sol and Sol Pro without restructuring their integration, and existing requests continue to work unchanged.

The GPT-6 family, including Sol Pro, introduces two additive behaviors compared with earlier OpenAI models: the model can keep working after it invokes one instead of pausing the turn for the application to return the result, and it accepts a change in reasoning effort partway through a conversation without invalidating the prompt cache. That mid-stream effort adjustment matters for long-running agentic sessions where a single hard turn benefits from deeper mode and surrounding turns can run lighter, and it removes the cache-invalidation penalty that previously forced developers to choose effort per request. For practical fit, Sol Pro targets workflows that need stronger reasoning than the base Sol tier on selected turns while staying inside the same tool-calling and conversation model as the rest of the GPT-6 line.