OpenAI has reduced prices for two GPT-5.6 models and introduced a faster processing option for its most capable model as the company seeks to make advanced AI workloads more economical at scale.

Starting July 30, GPT-5.6 Luna will cost 80% less while GPT-5.6 Terra will cost 20% less, according to an announcement from OpenAI. The company said the reductions reflect improvements across its models, inference infrastructure, production software, and context management systems.

Luna, the fastest and lowest cost model in the GPT-5.6 family, is now priced at $0.20 per million input tokens and $1.20 per million output tokens. Terra now costs $2 per million input tokens and $12 per million output tokens.

Pricing for GPT-5.6 Sol remains unchanged. The company also introduced Fast mode for Sol, replacing its previous Priority Processing service.

Fast mode can deliver up to 2.5× faster performance than Standard processing without changing the model’s intelligence, according to OpenAI. It costs twice the Standard processing rate, bringing Sol pricing under Fast mode to $10 per million input tokens and $60 per million output tokens.

Existing API requests using the priority service tier will automatically continue working, while developers can also select the new fast service tier. OpenAI said the change is backward compatible and does not require customers to update existing integrations.

The company positioned Luna as an option for large scale tasks including document analysis, customer interaction classification, routine coding, testing, and agent workflows involving tools and multiple steps.

OpenAI said Luna can deliver performance comparable to models considered frontier class a year ago at roughly 6 cents per dollar spent on each task and at nearly nine times the speed. 

The company also said Luna outperformed Fable 5 on its Agents’ Last Exam evaluation at an estimated cost per task nearly 99% lower.

OpenAI attributed part of the efficiency gains to work performed by GPT-5.6 Sol itself. Within a process supervised by humans, Sol rewrote production kernels, conducted experiments to improve token generation, and monitored training operations.

The kernel optimizations reduced the total cost of serving the model by 20%, while experiments conducted by Sol improved token generation efficiency by more than 15%, the company said.

Terra and Luna remain available through ChatGPT Work, Codex, and the OpenAI API. Free and Go subscribers can access Terra in ChatGPT Work and Codex, while Plus, Pro, Business, and Enterprise subscribers can select Terra or Luna.

OpenAI said ChatGPT and Codex subscription prices and overall quota budgets will remain unchanged. However, using Terra and Luna will now consume fewer credits under paid subscriptions.