🔍 Read the full analysis: The Fine Print Behind OpenAI And Anthropic’s AI Price Cuts on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI and Anthropic have sharply reduced prices for several newer AI models since June, with some input and output token rates falling by as much as 80% on OpenAI’s Luna line. Separately reported customer discounts and changing plan limits mean headline prices do not show the full cost: buyers need to check model performance, usage allowances, contract terms and credits.
OpenAI and Anthropic have lowered prices for several AI models since June, as a report published this weekend described OpenAI offering customer-specific discounts to win business from Anthropic. The reported deals point to growing competition for enterprise customers, but the savings figures vary by customer and do not represent across-the-board reductions in vendors’ list prices.
According to the source report’s review of vendor pricing pages and announcements, OpenAI’s Sol model line fell from $5 per million input tokens and $30 per million output tokens in June to $2 and $10 for GPT-6.1 Sol in September. The Luna line moved from $1 and $6 to 10 cents and 50 cents per million input and output tokens. These are comparisons across model versions, rather than evidence that the same model’s price fell while its capabilities stayed constant.
Anthropic also reduced prices on newer models. Its Opus line moved from $5 per million input tokens and $25 per million output tokens for Opus 5 in July to $4 and $20 for Opus 5.5. Anthropic also lowered the price of reading cached tokens for some models. Cache pricing can matter particularly for AI agents that repeatedly process conversation history, although the portion of total spending attributable to cached tokens depends on the workload.
A separate account of customer negotiations appeared in The Information on October 24, according to the supplied source. It described OpenAI discounts to attract customers, including a reported discount of more than 30 percent on a Replit contract worth over $100 million a year, and sub-$7 seats for a U.S. bank. The source also reports that DataCamp CEO Jonathan Cornelissen expects OpenAI’s offer to reduce his company’s total AI costs by 80% to 90%. That is an estimate of savings against DataCamp’s existing setup, not a 90% reduction from OpenAI’s public list price.
Infographic 01 · The price slide
Four months, a fraction of the price.
What a million output tokens cost on each vendor’s price list, from the June previews to today. Bars show the price; shorter is cheaper.
OpenAI · output price per 1M tokens
List prices, standard tier
Anthropic · output price per 1M tokens
List prices, standard tier
Same scale in both panels. Fable 5.1 kept its $50 output price and cut cache reads instead.
The quieter cut: cache reads
matters most for agents
Cache-read price per 1M tokens, same scale in all three.
Woolf estimates cached tokens make up 96–99% of token spend in agentic coding. Anthropic says Opus 5.5 costs 40% less than Opus 5 and Fable 5.1 about 25% (up to about 45%) less than Fable 5 on typical workloads; those are vendor claims. Most rows compare a successor model with its predecessor. Anthropic’s newer tokenizer produces about 30% more tokens for the same text, so per-token prices are not directly comparable across vendors.
Sources: OpenAI API pricing (26 Oct 2026); OpenAI announcements (26 Jun–29 Sep 2026); Anthropic API pricing (26 Oct 2026); Anthropic announcements (30 Jun–22 Sep 2026); Vercel AI Gateway index (Oct 2026); Max Woolf (25 Oct 2026).
Powered by Thorsten Meyer AI
How Lower Prices Change Buying Power
For organizations paying for large volumes of model use, lower rates and private discounts can materially affect the cost of running coding tools, assistants and other AI services. The reported customer moves suggest that price is influencing where some work gets routed: Rippling CEO Parker Conrad said OpenAI’s share of his company’s coding work rose from 16% in May to 52%, while Anthropic’s share fell from 57% to 18%. Those figures describe one company’s usage, not a market-wide shift.
Yet a lower price per token does not automatically mean a cheaper or better outcome. A model that needs more calls, performs less reliably on a particular task, or requires extra human review may cost more overall. Subscription limits also affect the calculation. The source says OpenAI reduced usage included in new $200 Pro subscriptions without changing the monthly price, while Anthropic added capped, expiring API credits to some plans. Buyers comparing offers need to measure actual workload costs and check the limits attached to each plan.
The competition may give customers more room to negotiate, especially when they can move workloads between providers. But savings described in company interviews or deal reporting are individual outcomes. They should not be treated as a standard discount available to every customer or as a guarantee of future costs.
Popular right now
TCL 85 Designer Series QD-Mini LED TV (Model 85A400U) with Google Smart TV
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
Breville Bambino Espresso Machine, Brushed Stainless Steel
As an affiliate, we earn on qualifying purchases.
A Four-Month Run of Price Changes
The sequence began with OpenAI’s June pricing for GPT-5.6 Sol and Luna. Anthropic introduced Sonnet 5 at $2 per million input tokens and $10 per million output tokens at the end of June, then made that introductory pricing permanent in August, according to the supplied account. OpenAI cut Luna’s rates in July and reduced Sol’s rates temporarily in August before September launches brought further price changes.
In September, Anthropic’s Fable 5.1 reduced its cache-read price, while OpenAI introduced GPT-6 Sol and Luna at lower rates. OpenAI then cut GPT-6.1 Sol’s cache price at the end of the month. Data scientist Max Woolf published a price timeline on October 25; the source says it checked his figures against vendor announcements and pricing pages retrieved October 26. Most comparisons involve successor models, so price movements alone do not establish that customers receive the same performance at a lower cost.
Anthropic has attributed additional workload savings to caching and model efficiency. The source reports Anthropic’s estimate that Fable 5.1 costs around 25% less on typical workloads and up to around 45% less on highly agentic workloads, and that Opus 5.5 costs 40% less than Opus 5 on typical workloads. These are company estimates, not independent measurements; results will depend on how customers use the models.
“OpenAI’s share of the company’s coding work rose from 16% in May to 52%, while Anthropic’s fell from 57% to 18%.”
— Parker Conrad, Rippling CEO, as reported in the supplied source
The Limits Behind Each Price
The available reporting does not disclose the full terms of the customer contracts, including how long discounts last, what usage commitments apply, or whether the reported rates are available to other buyers. The Information’s article is paywalled, and the supplied account says it relied on its public summary and two secondary write-ups. Deal details therefore remain partial.
It is also unclear how broadly the reported discounts are being offered, and whether customers can maintain them as workloads or contract periods change. The exact basis for every claimed savings estimate is not supplied. In addition, model performance and total costs can differ by task; vendor claims about savings on typical or agentic workloads are not independent, apples-to-apples tests for every buyer.
Plan rules add another uncertainty. The source says Anthropic’s Enterprise plan requires 20 seats, which could put it beyond the reach of many of the small advisory firms Schwab had promised access to. It also reports new caps and expiry dates for some Anthropic API credits, and reduced usage in new OpenAI Pro subscriptions. The source does not detail every affected allowance or how existing subscribers are treated.
What Buyers Should Check Next
Companies considering a change should compare offers using their own workloads, including input and output tokens, cached-token use, throughput, reliability and any human review required. They should ask vendors to spell out discount duration, renewal pricing, minimum seats, usage caps, credit expiry and overage charges, then test successor models against the tasks they currently run. Routing work across more than one provider can give buyers a comparison point, but it also requires checking technical and contractual constraints.
Individual users should watch included usage, not just the monthly fee. A subscription’s price can stay constant even when its allowance changes, while a lower-cost model may not be interchangeable with the model a user relied on before. The next meaningful evidence will be clearer contract terms and customer results over time, including whether reported savings hold as usage grows and promotional periods end.
No further contract details or broadly applicable discount schedule are provided in the source material. Until those are available, the reported offers should be treated as examples of negotiations, while public price pages and plan terms remain the more useful starting point for comparing options.
Key Questions
Did OpenAI cut its prices by 90%?
No general 90% list-price cut is reported. The 80% to 90% figure is DataCamp CEO Jonathan Cornelissen’s estimate of the total savings his company expects from OpenAI’s offer compared with its existing setup. The source separately reports a discount of more than 30% on a Replit contract.
How much did OpenAI’s model prices fall?
In the source’s comparison of vendor prices, OpenAI’s Sol line moved from $5 per million input tokens and $30 per million output tokens in June to $2 and $10 for GPT-6.1 Sol in September. Luna moved from $1 and $6 to 10 cents and 50 cents. These are prices for different model versions, not necessarily like-for-like comparisons.
Are Anthropic’s models also getting cheaper?
Yes. The source reports Opus pricing falling from $5 per million input tokens and $25 per million output tokens for Opus 5 to $4 and $20 for Opus 5.5. Anthropic also reduced cache-read rates for certain models and has made workload-savings claims that depend on usage patterns.
What should businesses compare beyond token prices?
They should check model performance on their own tasks, total usage, cached-token rates, subscription allowances, seat minimums, credit expiry, overage charges and the length of any negotiated discount. The monthly fee or headline token rate alone may not show total cost.
Source: ThorstenMeyerAI.com
