GPT-5.6 Sol sits at the top of the new GPT-5.6 family, released for general availability on July 9, 2026 after a limited preview, with Terra and Luna positioned as the balanced and cost-efficient siblings. OpenAI positions Sol as a frontier model aimed at demanding professional work: coding agents, long-running research, computer use, cybersecurity, science, and design tasks that require coordinating several tools. The launch messaging emphasizes "more intelligence from every token," framing Sol as an efficiency-first flagship designed to deliver more successful work per dollar rather than simply scaling raw capability. Practical fit centers on teams that need a single model able to handle hard, multi-stage jobs end to end, from inspecting and refining outputs to handing back ready-to-use results.

The model introduces an "ultra" capability setting that coordinates multiple agents across parallel workstreams to accelerate the hardest tasks, a notable step beyond single-pass inference. On Agents' Last Exam, a long-horizon evaluation across 55 professional fields, Sol sets a new high score of 53.6, leading the adaptive reasoning comparison by a wide margin and still beating the rival at medium reasoning with substantially lower token spend. OpenAI highlights stronger computer use and design judgment alongside state-of-the-art results in coding, knowledge work, cybersecurity, and science, positioning Sol as the most polished collaborator in the family. Together, these signals point to a model intended for organizations that want frontier reasoning on extended workflows without paying frontier-plus prices for routine turns.