Released less than three months after the previous version, these models represent a significant jump in reliability and cost efficiency for the company's AI infrastructure.
The release signals a shift in the competitive landscape as OpenAI prioritizes lower operating costs to compete with rivals like Anthropic.
GPT-6 Luna, the high-efficiency model, now matches the performance of the previous high-tier Sol model while costing only 1% of its price.
This reduction in capital expenditure is critical for scaling AI services, as it allows the company to extract more processing power from its existing data centers and provides a significant price advantage for businesses using their API (application programming interface).
Beyond performance gains, OpenAI has updated the models to use more natural language with less jargon and improved "alignment," a technical term for ensuring the AI's actions match human intent.
The update also introduces more efficient prompt caching—storing previously processed instructions—which can reduce costs by up to 90% for developers running long-term digital agents.
While these savings directly benefit API users, the company has not yet confirmed if or how these efficiencies will change pricing for individual subscribers on ChatGPT Plus or Pro plans.