OpenAI just made its cheapest AI models a whole lot cheaper. The company announced on July 30 that it’s cutting prices on its smaller GPT-5.6 models, with the entry-level Luna seeing an 80% reduction and the mid-tier Terra dropping 20%.
The timing is not subtle. Corporate America has been loudly questioning whether its AI spending is actually paying off, and a wave of lower-priced Chinese AI products has been eating into the market from below.
The numbers behind the cuts
The Luna model, OpenAI’s entry-level option in the GPT-5.6 family, saw the most dramatic reduction. Input costs dropped from $1 to $0.20 per million tokens. Output costs fell from $6 to $1.20 per million tokens. That’s an 80% haircut across the board.
The mid-tier Terra model got a gentler trim. Input pricing moved from $2.50 to $2 per million tokens, while output costs decreased from $15 to $12 per million tokens. A 20% reduction.
And then there’s the flagship Sol model, which stayed right where it was: $5 for input and $30 for output per million tokens.
Why this matters beyond Silicon Valley
This price cut didn’t happen in a vacuum. Just 11 days earlier, around July 19, OpenAI had already reduced GPT-4 Turbo API pricing by 20%. Two rounds of cuts in under two weeks signals something more than generosity. It signals pressure.
Chinese AI competitors have been offering comparable capabilities at significantly lower price points. The broader corporate scrutiny over AI spending adds another layer, with CFOs reviewing the return on AI integrations.
OpenAI’s pitch is essentially that its smaller models have gotten good enough to handle tasks that previously required the expensive flagship.