OpenAI Releases GPT-6 Sol and Luna With 50% Cheaper API Pricing

OpenAI has released two new models in its GPT-6 family, GPT-6 Sol and GPT-6 Luna, both available immediately through the OpenAI API under the identifiers gpt-6-sol and gpt-6-luna. The two models sit below GPT-6 Astra, the flagship tier that launched earlier this month, and were trained using methods similar to Astra's in order to bring its advances to faster, more affordable models. Sol is positioned for complex coding and professional tasks at lower cost, while Luna targets fast, high-volume everyday work. OpenAI says improved caching and inference efficiency allow it to serve both models more cheaply, and it is cutting Sol and Luna API prices by 50% against their GPT-5.6 promotional pricing. The GPT-6 family now spans three tiers: Astra for the hardest workloads, Sol for professional and coding tasks, and Luna for high-volume routine use. Both new models are API-only, meaning no weights are available for self-hosting. Deccan Chronicle reported the expansion of the GPT-6 lineup with the cheaper Sol and Luna models.

Global Impact

Economically, halving mid-tier API pricing compresses the revenue-per-token assumptions of rival labs and cloud providers, accelerating a price war that favors large-scale deployers over resellers. Technologically, pushing Astra-class training methods into cheaper tiers widens access to capable coding and agentic models, which should speed enterprise adoption and shift competitive pressure toward tooling and integration layers.