On July 30th, OpenAI reduced the price of GPT-5.6, but not across all tiers: the entry-level Luna saw an 80% reduction in cost per task, while the mid-tier Terra only experienced a 20% decrease. This stark difference in treatment for the same generation of models was attributed by the company to service efficiency improvements(the same computational power can handle more requests, spreading costs thinner).

Think of these three tiers of models as three tiers of buffet套餐: entry-level, mid-tier, and premium. This announcement says "price reduction," so you think everything's on sale, but look closer—the entry-level is slashed by 80%, while the mid-tier only gets a 20% cut, keeping it in that same middle ground. The analogy ends here, but the more crucial point is: whether the food is good or worth it doesn't depend on how much it’s discounted, but on what you can get at the same price in other tiers.
Official OpenAI visual for the GPT-5.6 price update
Official visual from OpenAI's GPT-5.6 pricing post · Source: openai.com
Numbers

First, Understand the Numbers: Which Tier Got What Discount

The adjustment, effective July 30th on the OpenAI API(the plug apps use to call the AI), only changed two tiers of pricing. The input/output prices for Luna dropped from $1.00/$6.00 per million tokens(the units AI bill by) to $0.20/$1.20—representing an 80% reduction in cost per task.

Terra decreased from $2.50/$15.00 to $2.00/$12.00, a 20% reduction in cost per task. In other words, the bulk of the price cuts were concentrated on the cheapest tier.

−80%
Luna's Cost Per Task Reduction
Input/output prices dropped from $1/$6 per million tokens to $0.2/$1.2. Source: OpenAI official, Artificial Analysis.
$0.20
Luna's Price Per Million Input Tokens
New input price after reduction, with output price similarly reduced to $1.20. Source: Official OpenAI pricing.
−20%
Terra's Cost Per Task Reduction
From $2.5/$15 to $2/$12, a much smaller decrease than Luna's. Source: OpenAI official, Artificial Analysis.
Judgment

Lower Price Doesn't Always Mean Better Value

This is the point most easily overlooked when dazzled by the price reduction numbers. The independent assessment from third-party evaluation firm Artificial Analysis is clear: even after the 20% reduction, Terra still lags behind Luna and Sol in the "intelligence level vs. cost per task" comparison chart.

It hasn't entered what is known as the value frontier(the group of options that are not comprehensively surpassed in both "how smart" and "how much" dimensions). In short, for the same amount of money, other tiers can give you smarter results, leaving Terra stuck in a middle ground.

Why Luna Was Cut to the Bone and Terra Was Spared

The focus of price reductions often follows where market competition is fiercest. The entry-level is currently the most competitive battlefield in the price war, and slashing the entry price to the bare minimum can directly deter those who "just need enough and only the cheapest."

Action

So, Should You Follow This Price Reduction?

The conclusion is not "switch because of the reduction," but rather, first clarify which category your work falls into:

Before Following the Price Reduction, Ask These Three Questions
1

First, categorize: For tasks that are high-volume, simple in logic, and "good enough," Luna is now extremely cheap and should be prioritized for switching.

2

Don't just look at the listed price: For tasks requiring stronger reasoning, test Luna/Terra/Sol with your actual tasks and choose based on "who gives better results for the same money," not who cuts prices more.

3

Pay attention to the details: The official price reduction is calculated based on API token prices, but your actual bill is also affected by context length, caching, and concurrency. Before migrating, run a small-scale test to reconcile the numbers.

This price reduction is precisely targeted at the cheaper tier, not a full-scale clearance sale. Luna has brought the "good enough" price to rock bottom, making it worth considering for high-volume tasks; but when you see the word "reduction," don't rush—first, ask yourself whether you want to get full or eat well.

This article is based on primary sources and cross-checked reporting. Vendor-published figures are the vendors' own and have not been independently verified unless noted.