OpenAI just fired the loudest shot yet in the AI price war. On July 30, 2026, it cut prices across its GPT-5.6 family, and the headline is hard to ignore: the light model, Luna, became 80 percent cheaper overnight. For any business running AI at scale, a cut that size rewrites the math on what is worth automating.
Luna’s price dropped from 1 dollar and 6 dollars to just 0.20 and 1.20 dollars per million input and output tokens. The mid-tier model, Terra, fell 20 percent, and only the flagship, Sol, held its price. OpenAI said the savings came from efficiency gains in its own infrastructure rather than a loss-leading land grab, though the competitive timing is impossible to miss.
What actually changed
GPT-5.6 comes in three tiers. Sol is the flagship you reach for on the hardest reasoning. Terra is the balanced middle. Luna is the light, fast, cheap model built for high-volume, everyday tasks: classifying messages, extracting data, drafting replies, tagging tickets. That bottom tier is where most production traffic actually lives, because most real work is not frontier-hard, it is just voluminous. Cutting Luna by 80 percent is therefore a cut aimed squarely at the workloads businesses run millions of times a day.
Why an 80 percent cut matters
Token cost is the main variable cost of putting AI into a live product or workflow. As long as it stays high, there is a line under which automation is not worth it, and countless small, repetitive tasks sit just below that line. Drop the price of the capable-but-cheap tier by 80 percent and that line moves sharply. Work that was too marginal to automate last month is comfortably profitable this month.
It is also a competitive move whether OpenAI frames it that way or not. When one major lab makes its cheap tier this cheap, rivals face pressure to match, and the price of everyday intelligence keeps falling. That is great news for buyers and brutal for anyone whose business model assumed AI would stay expensive.
What it means for your business
The practical takeaway is that cost is quietly disappearing as a reason not to automate. The bottleneck is shifting from whether you can afford to run something to whether you have a system that turns it into results. If you shelved an idea a few quarters ago because the token bill did not pencil out, it is worth pulling back off the shelf.
Think about the high-volume, low-glamour work in your operation: qualifying inbound leads, triaging support tickets, cleaning and enriching data, summarizing calls, drafting first-pass replies. At Luna’s new price, running those through AI at scale costs a rounding error compared to a year ago.
For sales teams tired of cold leads, slow customer responses, and manual processes, Dapta is the ultimate tool.
Dapta is the leading platform for creating AI sales agents specifically designed to increase inbound lead conversion. Respond to your leads in less than a minute with voice AI and WhatsApp that converts.
If you want your team to sell more while AI handles the complex stuff, you have to try it.
The companies that win the next stretch will not be the ones with access to the cheapest model, because everyone will have that. They will be the ones who already built the plumbing to point cheap, capable AI at a specific, repeatable process and let it run. Price was the excuse; now it is mostly gone.
OpenAI kicking off a price war is a signal that the cost of intelligence is on a steep downward curve. For teams in LATAM and global markets watching every line item, that is not an abstract milestone. It is an invitation to automate the things you assumed were too expensive to bother with.
