OpenAI’s cheapest model line, Luna, cost $6 per million output tokens when GPT-5.6 Luna reached the API on 9 July; GPT-6 Luna now costs $0.50, a 92% fall, according to OpenAI’s API changelog and pricing page. Terra fell from $15 to $5 and Sol from $30 to $10, both down 67%. Those are the figures Social Capital posted on X on 2 November, and they check out.
The cuts were not equal. On 9 July Sol cost 5 times Luna per token, on both input and output; it now costs 20 times as much, by our arithmetic from the same price tables. Compounded over the 116 days, Luna’s output price fell about 48% a month and Terra’s and Sol’s about 25% a month, our arithmetic.
Business Pill · The Jevons Paradox Explained
A short Business Engineer explainer on the Jevons paradox: when a resource gets cheaper to use, total use can rise. It teaches the general idea only and says nothing about any company in this piece.
The key insight: By our arithmetic from OpenAI’s price tables, Sol cost 5 times Luna per token on 9 July and costs 20 times Luna now, on both input and output: the bottom tier fell 92% while the middle and top of the Luna-Terra-Sol line fell 67%.
We rebuilt the timeline from OpenAI’s changelog and from Wayback Machine copies of its pricing page, taken almost daily from 9 July. All prices are Standard tier, per million tokens, for prompts of up to 272K input tokens.
OpenAI released the GPT-5.6 family on 9 July. The archived pricing page from that day lists GPT-5.6 Sol at $5 input and $30 output, GPT-5.6 Terra at $2.50 and $15, and GPT-5.6 Luna at $1 and $6.
Three weeks later the changelog said “GPT-5.6 Luna costs 80% less, while GPT-5.6 Terra costs 20% less”, and the pricing page moved Luna to $0.20 and $1.20 and Terra to $2 and $12 from 30 July.
Sol followed on 21 August, dropping to $4 input and $20 output, which OpenAI describes as “20% lower input pricing and 33% lower output pricing”. It calls this promotional pricing, “available at least through November 21, 2026”.
Then GPT-6 Astra arrived on 3 September, which OpenAI calls “our most capable model”. The pricing page lists it at $10 input and $50 output, above Sol.
The GPT-6 generation reset the lower lines on 22 September: GPT-6 Sol launched at $2 input and $10 output, and GPT-6 Luna at $0.10 and $0.50. A week later, on 29 September, GPT-6.1 Sol kept the $2 and $10 price. From the archived copy of 23 September until GPT-6.1 Terra arrived, the flagship price table listed no Terra model; GPT-5.6 Terra’s own model page still gives $12 output.
Terra was last to move. GPT-6.1 Terra launched on 26 October; its model page and the pricing page list $1 input and $5 output, the price we reported for GPT-6.1 Terra on Amazon Bedrock earlier today.
Standard-tier output price per 1 million tokens, prompts of up to 272K input tokens, for the newest Luna, Terra and Sol models on OpenAI’s flagship price table, 9 July to 2 November 2026. Dates from OpenAI’s API changelog; prices from its pricing page and Wayback Machine copies of it. GPT-5.6 Sol’s $20 is promotional pricing. The 5x and 20x ratios are our arithmetic.Animated: Standard-tier output price of OpenAI’s newest Luna model, US$ per 1 million tokens, weekly readings from 9 July to 22 October 2026 (unchanged to 2 November). Steps dated from OpenAI’s API changelog and checked against Wayback Machine copies of its pricing page; the -92% is our arithmetic. Chart: FourWeekMBA.
What Was Posted on X
Since July, the price of OpenAIโs Luna models has fallen 92%, from $6 to $0.50 per million output tokens.
Additionally, its Terra and Sol models are both down 67%.
Social Capital posted this chart on 2 November: Luna output down 92% since July ($6 to $0.50), Terra and Sol down 67%. We rebuilt every number from OpenAI’s changelog, its pricing page and 96 archived copies of that page: all six dollar figures match. Each fall spans a newer model; on the same GPT-5.6 models the cuts were 80%, 20% and 33% (Sol’s is promotional), and GPT-6 Astra at $50 now sits above Sol, outside the chart.
Where Our Figures Differ From Social Capital’s
Social Capital’s chart compares the GPT-5.6 launch prices of 9 July with today’s list prices, and every dollar figure on it matches OpenAI’s pages. Its 92% is 91.7% unrounded, and its 67% is 66.7%.
The difference is in what is being compared. Each drop spans a newer model, not a cut to the same one. On the same model, the cuts were smaller: GPT-5.6 Luna fell 80%, GPT-5.6 Terra 20% and GPT-5.6 Sol 33%, and OpenAI’s model pages still list those prices. Sol’s cut is promotional.
Input prices fell less than output prices on all three lines: Luna input is down 90%, from $1 to $0.10, while Terra and Sol input are down 60%, from $2.50 to $1 and from $5 to $2, by our arithmetic.
Social Capital’s chart describes Luna, Terra and Sol as OpenAI’s tiers “from smallest to largest”. Since 3 September, GPT-6 Astra has sat above Sol on OpenAI’s price table at $50 per million output tokens, a price not on the chart.
What the Archive Shows
We read 96 Wayback Machine copies of OpenAI’s pricing page taken between 9 July and 2 November, plus the live page. Each price change above shows up in the archived copies within a day of its changelog date, and no copy we read shows a price for these tiers that the changelog does not explain. For 21 of the 117 days we have no copy, so a price that lasted less than a day or two on those dates would not appear in our record.
As we read it, the business lever is price: across the Luna, Terra and Sol lines, OpenAI lowered the bottom tier fastest.
The Structural Read
As we read it, the July-to-November fall is two moves stacked: price cuts on the GPT-5.6 models already on sale (80% for Luna, 20% for Terra, 33% for Sol on promotion), then new models listed lower still.
As we read it, the ladder also grew a rung at the top: GPT-6 Astra, listed at $50 per million output tokens since 3 September, sits above Sol, so the cheapest and dearest prices on OpenAI’s flagship table moved apart.
As we read it, a per-token price is not a per-task cost: the successor models are different models, and OpenAI’s pages do not say how many tokens each uses for the same job.
OpenAI API changelog, 30 July 2026
“GPT-5.6 Luna costs 80% less, while GPT-5.6 Terra costs 20% less”
Three Implications
FOR BUILDERS ON LUNA GPT-6 Luna lists at $0.10 input and $0.50 output, against $1 and $6 for GPT-5.6 Luna at launch on 9 July, according to OpenAI’s pricing page and its archived copy.
FOR SOL USERS GPT-6.1 Sol lists at $2 and $10; GPT-5.6 Sol’s $4 and $20 is promotional pricing that OpenAI says runs at least through November 21, 2026.
FOR TERRA USERS GPT-6.1 Terra lists at $1 and $5 on OpenAI’s API, the same global rate Amazon lists on Bedrock, as we reported earlier today.
The framework describes the token this way: “It is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts.”
As we read it, OpenAI’s price timeline is the third of those units moving: the price the lab charges per token fell by two-thirds or more on every Luna, Terra and Sol line in 116 days. The framework is the owner’s argument and does not discuss these price changes.
OpenAI’s pages give list prices, not what customers pay after discounts, and not usage or revenue by tier. OpenAI has not said whether GPT-5.6 Sol’s promotional price ends on 21 November. The successor models are different models, so a lower price per token is not the same as a lower cost per task. The monthly rates and the 5x and 20x ratios are our arithmetic.
The days without an archived copy are 12, 13, 20, 26 and 27 July; 3, 8, 9, 10, 16, 17, 18, 19, 23 and 30 August; 3, 13, 16, 19 and 28 September; and 25 October. We did not contact OpenAI or Social Capital.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Since 9 July, OpenAI’s Luna output price has fallen from $6 to $0.50 per million tokens, and Terra’s and Sol’s by two-thirds, according to OpenAI’s changelog and pricing page. Social Capital’s figures are correct; each drop spans a newer model, and Sol now costs 20 times Luna, up from 5 times.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Dates and price changes come from OpenAI’s API changelog; July prices from a Wayback Machine copy of OpenAI’s pricing page taken on 9 July; current prices from OpenAI’s pricing and model pages, read on 2 November. Percentages, ratios and monthly rates are our arithmetic. We did not contact OpenAI or Social Capital.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
