OpenAI has lowered the price of two GPT-5.6 models and introduced a faster processing option for its most capable model, changing the cost and latency choices available to API customers. The 30 July update applies to GPT-5.6 Luna and Terra pricing, while a new Fast mode gives developers an alternative to Standard processing for GPT-5.6 Sol.
The company says Luna now costs 80 per cent less and Terra costs 20 per cent less. OpenAI has positioned Luna as the fastest and least expensive member of the GPT-5.6 family, intended for cost-sensitive, high-volume workloads. Terra sits between Luna and Sol, balancing capability, speed and cost for everyday production tasks.
New GPT-5.6 Luna and Terra API prices
From 30 July, GPT-5.6 Luna is priced at US$0.20 per million input tokens, US$0.02 per million cached input tokens and US$1.20 per million output tokens. GPT-5.6 Terra is priced at US$2 per million input tokens, US$0.20 per million cached input tokens and US$12 per million output tokens.
Those figures make the price difference between the two models a straightforward tenfold step across standard input, cached input and output tokens. For developers, that creates a clearer decision point: Luna may suit routine or tightly specified work where throughput and cost are the main constraints, while Terra may be appropriate when a task benefits from a stronger balance of reasoning and economy.
The headline token rates do not cover every usage pattern. OpenAI's model documentation says prompts above 272,000 input tokens are charged at twice the input rate and 1.5 times the output rate for the entire request. Cache writes are billed at 1.25 times the uncached input-token rate. Teams working with very long contexts should therefore model their actual request sizes rather than estimating spend from the base rates alone.
Fast mode replaces Priority Processing
OpenAI is also replacing its Priority Processing option with Fast mode in the API. For GPT-5.6 Sol, the company says Fast mode can deliver up to 2.5 times the speed of Standard processing at twice the price, without changing the model's intelligence. The claim describes a maximum improvement rather than a guaranteed latency for every request, so production users will still need to measure performance against their own workloads and traffic patterns.
The change is designed to be backward compatible. Existing API requests tagged for priority processing will automatically use Fast mode, which should reduce the immediate migration burden for current Priority Processing customers. OpenAI also says the API option aligns with the /fast control in Codex.
Fast mode gives developers a separate lever from model selection. A workflow can use Luna or Terra to reduce the cost of repeatable work, reserve Sol for tasks that require its higher capability, and then pay the Fast mode premium only when response time is important. That structure may be useful for multi-stage agents, but it also increases the need for routing rules and evaluation data so that each stage uses an appropriate model and processing tier.
Availability across OpenAI products
GPT-5.6 Terra and Luna remain available through the OpenAI API, Codex and ChatGPT Work. OpenAI says Free and Go users can access Terra, while Plus, Pro, Business and Enterprise users can choose Terra and Luna. Subscription prices and quota budgets are unchanged; instead, usage of Terra and Luna will consume fewer credits in Codex and ChatGPT Work.
The company said the price changes would begin rolling out through AWS later on 30 July. That wording suggests customers using OpenAI models through AWS should confirm when the updated rates appear in their own billing path before relying on the lower cost in forecasts.
OpenAI's official model pages list the same 1,050,000-token context window and 128,000-token maximum output for both Luna and Terra. Both support reasoning and image input as well as text, while audio and video are not supported as model inputs. Their Responses API tool support includes web search, file search, code interpreter, hosted shell, computer use, MCP and structured tool workflows.
What the changes mean for buyers
The practical impact depends on cost per completed task, not token price alone. A less expensive model can still cost more if it requires longer prompts, more retries or additional verification. Conversely, a higher-priced model may be economical when it reaches an acceptable result with fewer steps. OpenAI cites internal efficiency improvements and customer examples, but organisations should treat those as company-reported results and run evaluations using their own prompts, tools and acceptance criteria.
The update does not introduce a new model family or change Sol's standard price. Instead, it adjusts the economics of the existing GPT-5.6 range and separates faster processing from the underlying model choice. For teams already using Luna, Terra or Priority Processing, the immediate work is to check billing, confirm routing behaviour and revisit model-selection thresholds. New projects can use the revised prices as a starting point, but should still account for long-context multipliers, cache-write charges and the Fast mode premium.