OpenAI has expanded its GPT-6 lineup with two lower-cost models, GPT-6 Sol and GPT-6 Luna, offering capabilities developed from the same methods used for its flagship GPT-6 Astra while significantly reducing API costs.
The launch follows the introduction of GPT-6 Astra earlier this month. OpenAI describes Astra as its most capable model for demanding projects, while Sol and Luna are designed to make advanced AI capabilities more accessible across different workloads and budgets.
GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens, representing a 50% reduction from the promotional pricing of its GPT-5.6 predecessor. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens.
OpenAI said the new models are designed for professional work, coding, automation and computer-use tasks. Sol is positioned for more complex workloads, while Luna is aimed at faster, high-volume tasks where cost efficiency is particularly important.
The company said Sol and Luna were trained using methods similar to those behind Astra, bringing improvements in professional work, factuality, coding, computer use and alignment to the lower-cost models. OpenAI also said Sol makes about half as many mistakes as its predecessor on its internal factuality evaluation.
India Edge Japan by Two Runs in Rain-Shortened T20I
OpenAI attributed the lower pricing to improvements in caching and inference efficiency. The company said those infrastructure gains allowed it to reduce API prices and pass the savings on to users and customers.
The new models are being rolled out across OpenAI’s products and API services. They are available in ChatGPT Work and Codex for eligible paid users, while GPT-6 Luna is also being made available to Free and Go users through the desktop app. Availability can vary during the rollout.
The launch comes as major AI companies increasingly focus on lowering the cost of running more capable models. The competition is also intensifying, with Anthropic releasing a new Opus model on the same day as OpenAI’s Sol and Luna announcement.
OpenAI has also faced scrutiny over the behaviour of AI agents and the broader challenge of keeping increasingly capable systems aligned with human oversight. The company has publicly disclosed several instances of concerning model behaviour during internal testing, underscoring the continuing safety challenges surrounding more autonomous AI systems.
With Astra remaining at the top end of OpenAI’s model lineup, Sol and Luna expand the GPT-6 family toward lower-cost applications, particularly coding, automation and other tasks that require repeated or large-scale AI usage.
