AI Just Got 80% Cheaper — What That Means for Your Business
On July 30, OpenAI cut the price of its GPT-5.6 Luna model by 80% in a single move — from $1.00 to $0.20 per million input tokens. Days later, the same model was set to become the default for ChatGPT’s free tier. OpenAI credits efficiency gains and, plainly, competitive pressure: Chinese labs like DeepSeek and Alibaba keep releasing capable models at commodity prices, and the market keeps responding.
If you run a business, this isn’t tech-industry inside baseball. It’s a line item on your P&L changing by an order of magnitude.
The math that changed
A year ago, a custom AI workflow — say, one that reads every inbound lead, qualifies it against your criteria, drafts a reply, and logs it to your CRM — carried real per-use costs. Enough that many small businesses reasonably concluded custom AI was for companies bigger than theirs.
That math is gone. Work that cost hundreds of dollars a month in model usage now costs a fraction of it. The 2026 Stanford AI Index reports that 88% of organizations now use AI in at least one business function — and the price war is a big part of why. The barrier was never whether AI could do the work. It was whether the work penciled out. It pencils out now.
For the automation projects we build — lead follow-up, document processing, report generation, customer-support drafting — model cost has quietly stopped being the constraint. The constraint is knowing which workflows are worth automating and building them so a person stays in control.
The catch: prices move in both directions
Here’s the part of the story that doesn’t make headlines. The same month OpenAI cut prices 80%, another major lab’s promotional pricing expires — with standard rates 50% higher on both input and output starting September 1. A workload that costs $2,000 a month on the promo rate becomes a $3,000 workload with zero change in usage.
Same summer. Same industry. One price fell off a cliff; another jumped 50%.
That’s the real lesson for business owners, and it’s the one we build around: never weld your business to a single AI model. Models are retiring faster than ever — three separate retirement dates landed in August alone. A system hard-wired to one vendor’s model inherits every price change and every retirement date that vendor decides on.
How to actually capture the savings
Three practical moves:
Build model-agnostic. Your automation should treat the AI model like a replaceable part — when a better or cheaper model ships, you swap it, not rebuild. Every system Tenvaro ships is built this way, because we’ve watched this movie before.
Revisit the “we priced it last year” list. If you scoped an AI project in 2025 and shelved it on cost, that estimate is stale. Re-run the numbers before your competitors do.
Track what your AI actually costs. Falling unit prices have a sneaky side effect: usage grows to fill the budget. This is a business-intelligence problem — a simple dashboard tying AI spend to outcomes (leads handled, documents processed, hours saved) tells you whether cheap AI is actually earning its keep.
The window is open
Cheap frontier AI is a moment, not a permanent condition — vendors are testing what the market bears in both directions. Businesses building AI fluency now, on flexible foundations, compound that advantage every month the price war continues.
Tenvaro builds AI automation and business intelligence for U.S. businesses — model-agnostic by design, with a human approval gate before anything important happens. If you shelved an AI project on cost, talk to us. The math changed.
Sources: Kraviona — AI News August 2026 · Digital Applied — AI Model Releases: August 2026 Tracker