On July 30, OpenAI did something every business owner running, or considering, AI automation should know about: they slashed the price of their GPT-5.6 Luna model by 80%.
Not a promotional rate. Not a trial period. A permanent reduction in what you pay per unit of AI output. Luna’s input token price dropped from $1 to $0.20 per million tokens; output dropped from $6 to $1.20. If you’re running document processing, customer query triage, or data classification through the API, that’s the same workflow at one-fifth the cost, with no code changes required.
This isn’t a story about OpenAI being generous. It’s a story about what happens when a market gets genuinely competitive.
Why OpenAI Cut Prices, and Why the Reason Matters More Than the Number
The timing isn’t subtle. In late July, Anthropic launched Claude Opus 5 at roughly $5 per million input tokens, with benchmark performance matching OpenAI’s top-tier model. Google shipped a faster, cheaper Gemini tier around the same window. Within days, OpenAI responded by cutting Luna’s price by 80%.
What you’re watching is frontier AI pricing behaving the way cloud computing did between 2010 and 2015: aggressive, structural, and unlikely to reverse. AWS, Azure, and Google Cloud slashed storage and compute prices repeatedly as competition intensified, and those prices never went back up. AI API pricing is now on a similar trajectory.
What This Actually Changes for a Business Evaluating AI
For businesses that shelved an AI automation project because the API costs didn’t pencil out, this is the moment to revisit it. The cost argument against AI automation is getting harder to sustain with every price cut like this one — not because AI got smarter, but because the unit economics of running it moved.
The task worth re-costing isn’t “AI” broadly. It’s the specific, well-defined workflow you priced out before: a document classification step, a first-pass customer query triage, a data extraction task that used to run at a cost that didn’t justify the automation. At one-fifth the price, the same task may now clear the bar it didn’t clear six months ago.
The Part Worth Being Careful About
A price cut on one model doesn’t mean every AI vendor’s pricing has moved, and it doesn’t mean every task suddenly makes sense to automate. The businesses that benefit are the ones that already know precisely which task they’d automate and roughly what it currently costs to run manually — not the ones chasing a general sense that “AI is cheaper now.” Re-costing a specific workflow takes an afternoon. Adopting AI broadly because prices fell somewhere is how a business ends up with tools nobody’s actually using.
What This Means for Singapore SMEs
Singapore’s SME AI adoption has roughly tripled in recent years, part of a digital economy IMDA’s 2025 Singapore Digital Economy Report puts at 18.6% of GDP. Falling frontier AI prices lower the barrier to that adoption curve directly — a document-heavy task that didn’t justify automation at last year’s API pricing might justify it now. The practical move is narrow: pick the one task you priced out before, re-run the numbers at current pricing, and test it before assuming the maths still says no.
Enterprise Singapore’s Productivity Solutions Grant continues to co-fund automation tooling built on this kind of infrastructure — worth checking what the PSG grant covers before ruling out a properly-scoped pilot on cost grounds.
About the Author
Keith Kwai is an independent consultant helping Singapore SMEs with digital transformation and AI adoption. He has 25 years of marketing and digital experience across global MNCs including Motorola, Singtel, Creative Technology, Epson, and Scholastic International. He builds agentic AI systems — not just advises on them. Connect at keithkwai.com or LinkedIn.
Sources: CNBC, OpenAI GPT-5.6 pricing coverage; IMDA Singapore Digital Economy Report 2025; Enterprise Singapore Productivity Solutions Grant (PSG) guidelines.