OpenAI has just dropped the GPT-5.6 family, and it’s a big deal. The lineup includes three models: Sol, the new flagship; Terra, a solid everyday workhorse; and Luna, the budget-friendly option. Sol is the star here, setting new benchmarks in coding, cybersecurity, and scientific reasoning while using way fewer tokens and costing less than competitors like Claude Fable 5. In fact, Sol’s max reasoning mode nearly ties Fable 5 on the Artificial Analysis Intelligence Index but does the job in 61% less time and at half the cost. There’s also a new ‘ultra’ setting that coordinates multiple AI agents to tackle complex tasks faster. Safety hasn’t been ignored either: OpenAI claims these models have their strongest safeguards yet, tested through extensive red teaming and automated checks. The coding prowess is impressive too — Sol’s top reasoning mode scored 80 on the Coding Agent Index, outperforming Fable 5 while using less than half the output tokens and costing about a third less. Why it matters: OpenAI is finally delivering on the promise of making frontier AI both more powerful and more accessible, challenging the industry to compete on efficiency rather than just raw capability.