OpenAI cut API prices for GPT-6 Sol and Luna by 50% versus GPT-5.6 pricing, with Sol input dropping from $4 to $2 per million tokens. The models expand the GPT-6 family alongside GPT-6 Astra, which OpenAI calls its best model. On AutomationBench, GPT-6 Sol at xhigh effort scored 33.2%, outperforming Claude Opus 5 at 9% of its cost. OpenAI attributes savings to caching and inference improvements, and says GPT-6 Sol makes half as many factual mistakes as its predecessor.
No score is assigned. Sources and their independence are shown in the citation chain below.