OpenAI Cuts GPT-5.6 Luna to $0.20 Per Million Input Tokens
By AgentRiot Editorial
OpenAI has cut GPT-5.6 Luna API input pricing to $0.20 per million tokens, cut Terra by 20%, and added Sol Fast Mode for latency-sensitive work.
OpenAI has turned GPT-5.6 Luna into a materially cheaper production option: the model now costs $0.20 per million input tokens and $1.20 per million output tokens. That is an 80% price cut, effective July 30.
The headline matters because Luna now sits at one tenth of GPT-5.6 Terra’s published rate. Terra costs $2 per million input tokens and $12 per million output tokens after its own 20% reduction. For a simple workload with 10 million input tokens and 1 million output tokens, the published API rates are $3.20 on Luna and $32 on Terra, excluding cached-token discounts, batch pricing, and any other platform charges.
That gap does not settle which model to use. It does make routing a much less academic exercise.
The decision is now about where quality changes the outcome
OpenAI describes Luna as its fastest and lowest-cost GPT-5.6 model, while Terra is positioned for everyday work. The useful question for an agent workflow is not whether a cheaper model can do every job. It is whether a given step actually benefits from a higher-cost model.
A well-bounded task, such as sorting documents, extracting structured fields, or implementing a change from a settled specification, may be a reasonable candidate for a lower-cost route. A task with unresolved requirements, a high cost of error, or a need for stronger judgment may not be.
OpenAI gives its own coding example: use Sol to resolve uncertainty and make a plan, then use Luna for well-specified implementation, test runs, and evaluation. That is product guidance from the vendor, not an independent benchmark. Still, it describes the practical effect of the new prices: teams can revisit the handoff points inside a workflow instead of treating every agent step as the same class of problem.
Sol now has a speed premium, not a price cut
OpenAI also replaced Priority Processing with Fast mode for GPT-5.6 Sol. The company says Fast mode delivers up to 2.5 times the speed of Standard processing at twice the Standard price, with no change in intelligence. Sol’s base price is unchanged.
Existing API requests tagged priority will automatically use Fast mode, according to OpenAI. The company says the offer also aligns with /fast in Codex.
This is a different proposition from the Luna and Terra reductions. Fast mode is for latency-sensitive work where waiting less is worth paying more. It is not a way to get Sol more cheaply.
Fewer credits, same subscription price
OpenAI says the lower Luna and Terra API prices also reduce how quickly usage is counted against paid Codex and ChatGPT Work subscriptions. Subscription prices and quota budgets are unchanged; the affected models consume fewer credits.
The company says Free and Go users can access Terra in ChatGPT Work and Codex. Plus, Pro, Business, and Enterprise users can choose Terra and Luna. Both models remain available through the API.
The price changes begin rolling out through AWS on July 30, OpenAI says. Before moving production traffic, teams should measure the actual token mix, test the lower-cost route against their quality threshold, and decide whether Fast mode’s latency premium pays for itself. The price cut is real. The correct routing decision remains workload-specific.