Advancing the price-performance frontier with GPT-5.6

OpenAI has updated its pricing and performance capabilities for the GPT-5.6 model family. The company is reducing costs for its Luna and Terra models while introducing faster processing speeds for its flagship Sol model in the API.

GPT-5.6 Luna is now 80% more affordable than previous configurations. This reduction makes it a viable option for high-volume tasks that require agents to manage multi-step workflows or handle large datasets. Terra, which serves as the balanced model for standard business operations, has seen a 20% price reduction. These changes are designed to allow businesses to deploy artificial intelligence across a broader set of daily tasks.

For users who prioritize speed, the new Fast mode replaces the previous Priority Processing offering for GPT-5.6 Sol. This mode provides up to 2.5 times faster speeds compared to standard processing. While this service comes at a higher cost, it allows for faster turnaround times on complex, time-sensitive workloads without any loss in the underlying intelligence of the model.

These efficiency gains are the result of improvements in model architecture and inference systems. By utilizing GPT-5.6 Sol to optimize production kernels and refine token generation, the technical team has reduced the costs associated with serving these models. This process creates a feedback loop where the models themselves help identify ways to operate more efficiently.

The updated pricing structure is now active for API customers, and changes are rolling out through AWS. These adjustments reflect a focus on matching specific workloads to the appropriate model, giving enterprises the ability to choose between high-end capability for complex planning and cost-effective processing for routine execution.