OpenAI Rolls Out GPT-5.6 and Cuts AI Model Prices by Up to 80%
OpenAI has launched its new GPT-5.6 model family and announced sweeping price reductions across two of its lower-cost variants, as the company seeks to make advanced artificial intelligence more affordable for developers and businesses handling high-volume workloads.
The GPT-5.6 family includes three main models: Sol, designed for the most demanding reasoning and professional tasks; Terra, positioned as a balanced model for everyday work; and Luna, the fastest and most affordable option in the series.
OpenAI said it has reduced the price of GPT-5.6 Luna by 80%, while cutting the cost of GPT-5.6 Terra by 20%. The reductions also affect how usage is calculated for paid subscribers using the models through ChatGPT Work and Codex.
The company is positioning Luna as a cost-efficient choice for businesses processing large volumes of requests, including customer support, document analysis, content operations and automated workflows. Terra, meanwhile, is aimed at organisations seeking stronger reasoning capabilities without the higher cost associated with the flagship Sol model.
GPT-5.6 has been developed to deliver more useful work with fewer tokens, improving the balance between model performance and operating costs. OpenAI said the gains stem from changes across model training, inference infrastructure, caching and workload management.
The company has also introduced a new Fast mode for GPT-5.6 Sol through its application programming interface, offering speeds of up to 2.5 times those available under standard processing. The faster service is priced at twice the standard rate.
The GPT-5.6 series supports advanced applications across coding, scientific research, cybersecurity, multimodal analysis and long-context tasks, while giving developers the flexibility to select different performance and pricing levels depending on the complexity of their workloads.
The latest pricing move highlights intensifying competition among major artificial intelligence companies, as model providers increasingly focus not only on performance but also on reducing the cost of deploying AI at scale.