OpenAI cuts GPT-5.6 Luna prices by 80% and speeds up Sol

OpenAI has reduced API prices for its GPT-5.6 Luna and Terra models, while introducing a faster processing option for GPT-5.6 Sol. The changes are designed to lower the cost of high-volume AI work and give customers faster responses when needed. OpenAI announces the changes in an official blog post.

GPT-5.6 Luna, the company’s lowest-cost and fastest model, now costs 80% less. Its API price is $0.20 per million input tokens and $1.20 per million output tokens. Terra, positioned for everyday tasks, receives a 20% price reduction. It now costs $2 per million input tokens and $12 per million output tokens.

OpenAI says the new prices also reduce how much Terra and Luna usage counts against paid quotas in Codex and ChatGPT Work. Subscription prices and total quota budgets remain unchanged.

Faster processing for complex work

The company is also replacing its API Priority Processing option with Fast mode. For GPT-5.6 Sol, Fast mode can deliver responses up to 2.5 times faster than Standard processing, according to OpenAI. It costs twice as much as Standard processing, but OpenAI says it does not alter the model’s intelligence. Existing API requests marked as priority will automatically move to Fast mode.

OpenAI attributes the price and speed changes to improvements in model design, inference systems, and agent workflows. The company says better hardware routing, more efficient token generation, and improved context management reduce the computing resources needed for a task.

The company also says Sol has assisted engineers with production software optimization and experiments aimed at improving token generation. OpenAI reports that this work reduced the end-to-end serving cost for Sol by 20% and increased token-generation efficiency by more than 15%.

For users, the revised lineup creates a clearer division of work: Sol for demanding tasks where speed or deeper reasoning matters, and Luna or Terra for routine, large-scale workflows such as document analysis, classifications, testing, and implementation.