OpenAI has announced a significant price reduction for two of its GPT-5.6 family of AI models , just twenty days after their official launch. The company has reduced the cost of GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20% , while for its flagship model, GPT-5.6 Sol, it has introduced a faster API option. These changes, which took effect on July 30, aim to improve the balance between cost, speed, and capacity, especially for companies handling large volumes of AI-powered automated tasks.
The company explained that the price reductions are due to optimizations across its entire software stack, including improved hardware routing, enhanced inference software, and smarter context caching algorithms. In fact, some of these improvements were driven by the models themselves: during supervised testing, GPT-5.6 Sol autonomously rewrote and optimized critical production software cores, reducing service overhead by 20% and improving token generation efficiency by more than 15%. This allows OpenAI to extract greater performance from its existing server clusters.
New prices for GPT-5.6 Luna and Terra
With the new pricing, GPT-5.6 Luna now costs $0,20 per million tokens in and $1,20 per million tokens out, down from $1 and $6, respectively. This model, the fastest and most cost-effective in the family, is designed for high-volume tasks such as customer interaction classification, massive document analysis, and background automation. OpenAI claims it offers performance comparable to models considered cutting-edge approximately a year ago, but at a fraction of the cost.
Meanwhile, GPT-5.6 Terra, designed for everyday work, is priced at $2 per million tokens invested and $12 per million tokens issued. Its approach seeks a balance between reasoning, speed, and cost for business assistants, programming, and workspace-related queries. The strategy allows for combining different models within the same process: for example, a company could use Sol to resolve uncertainties and develop a plan, and then assign Luna to implementation, testing, and results evaluation.
Fast mode: extra speed for GPT-5.6 Sol
OpenAI has also introduced Fast mode for GPT-5.6 Sol, replacing the previous Priority Processing in the API. This mode allows the flagship model to run up to 2,5 times faster than standard processing, although it costs twice as much to use. The company assures that the increased speed does not compromise the model's intelligence. Existing requests identified as priority will automatically be switched to Fast mode, maintaining compatibility with current implementations.
These pricing changes also affect the credit or time limits for users and organizations on ChatGPT Work and Codex, where using Luna and Terra will consume fewer credits without altering monthly subscription levels. OpenAI announced the news via its X account, stating that "with a commitment to pushing the boundaries of models in terms of cost efficiency, capacity, and speed," they have reduced prices. The price cut comes at a time of heightened competition in the sector, with the recent emergence of models like Kimi K3 from the Chinese startup Moonshot AI.
The new pricing has been in effect since July 30th, and while performance data is primarily based on internal OpenAI assessments and selected customers, the company is confident that savings will be sustained across diverse and large-scale workloads. With this strategy, OpenAI reinforces the trend of segmenting inference based on problem complexity, enabling organizations to accurately weigh financial costs against the computing power required for their operations.