Terra is also becoming 20% cheaper, while a new premium mode promises significantly faster performance for GPT‑5.6 Sol.

OpenAI is sharply lowering the cost of running two members of its GPT‑5.6 model family, making advanced artificial intelligence more affordable for businesses, developers, and high-volume automated workflows.

In a July 30 announcement, the company said it is reducing the price of GPT‑5.6 Luna by 80% and GPT‑5.6 Terra by 20%. OpenAI is also introducing a faster premium processing option for GPT‑5.6 Sol, its most capable model.

The changes arrive just three weeks after the GPT‑5.6 family became generally available.

The New GPT‑5.6 Prices

Starting July 30, API customers will pay:

• GPT‑5.6 Luna: $0.20 per million input tokens and $1.20 per million output tokens.
• GPT‑5.6 Terra: $2 per million input tokens and $12 per million output tokens.
• GPT‑5.6 Sol: Standard API pricing remains unchanged.

Tokens are the small units of text that AI models process when reading instructions and generating responses.

For individual users, it is important to understand that ChatGPT and Codex subscription prices are not being reduced. Existing quota budgets also remain unchanged. However, using Terra and Luna will now consume fewer credits, allowing subscribers to complete more work within their current plans.

Free and Go users can access Terra, while Plus, Pro, Business, and Enterprise subscribers can choose between Terra and Luna.

Why the Luna Price Cut Matters

The cost of a single AI interaction may seem small, but expenses can grow quickly when companies process thousands of documents, classify customer requests, run background automations, generate code, or operate AI agents throughout the day.

Luna is positioned as the fastest and most affordable GPT‑5.6 model. Unlike basic low-cost models, it can use tools and complete workflows involving multiple steps. That makes it suitable for routine coding, document processing, content organization, customer-service classification, testing, and other repeatable tasks.

OpenAI says Luna now delivers performance comparable to models considered frontier-class a year ago at roughly 6% of the estimated cost per task and nearly nine times the speed. Those comparisons are based on company-reported evaluations, and actual results will depend on the type and complexity of each workload.

Different Models for Different Jobs

The broader strategy behind GPT‑5.6 is not to use the most powerful model for every task.

A business could use Sol to analyze an uncertain problem and create a plan, then send clearly defined execution work to Luna. Terra can handle everyday professional tasks that require a stronger balance between intelligence, speed, and cost.

This approach allows companies to reserve premium AI performance for difficult decisions while using less expensive models for high-volume production work.

Sol Gets a New Fast Mode

OpenAI is also replacing its Priority Processing option with a new Fast mode for GPT‑5.6 Sol in the API.

The company says Fast mode can deliver responses up to 2.5 times faster than Standard processing while maintaining the same level of intelligence. The additional speed comes at twice the Standard API price.

Existing API requests marked as “priority” will automatically use Fast mode, so developers will not need to rebuild their current integrations.

How OpenAI Says It Lowered Costs

According to OpenAI, the savings come from improvements across model design, infrastructure, hardware routing, production software, and the way AI agents manage context during longer assignments.

The company also used GPT‑5.6 Sol during a human-led engineering process to optimize parts of the systems running its models. OpenAI says that work helped reduce end-to-end serving costs by 20%, while separate experiments improved token-generation efficiency by more than 15%.

These remain OpenAI-reported figures rather than independent performance tests.

The Bottom Line

The biggest change is not simply that one AI model became cheaper. OpenAI is making it more practical for companies to divide work across different levels of intelligence.

Sol can handle the most complex decisions. Terra can manage balanced everyday work. Luna can take on routine, high-volume execution at a much lower cost.

For businesses building AI into their daily operations, that flexibility could ultimately matter more than any single benchmark score.