🔍 Read the full analysis: Lower AI Prices, New Customers: OpenAI And Anthropic’s Pricing Battle on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI and Anthropic have cut several model prices since June, while OpenAI is also offering reported contract discounts to win enterprise customers. Companies cited in recent reporting are shifting some workloads, but the widely repeated claim of 80% to 90% savings describes one customer’s expected overall cost reduction, not a general OpenAI list-price discount.
OpenAI and Anthropic have lowered prices for several AI models since June, and new reporting says OpenAI is offering discounts to attract customers from Anthropic. The moves are changing costs and provider choices for businesses, although reported savings vary by customer and the frequently cited 80% to 90% figure is one executive’s estimate of total expected savings, not a general discount off OpenAI’s published prices.
Pricing pages and vendor announcements reviewed by ThorstenMeyerAI.com on October 26 show sharp reductions across some models. OpenAI’s Sol model line moved from $5 input and $30 output per million tokens in June to $2 and $10 for GPT-6.1 Sol in September. Its smaller Luna line fell from $1 and $6 to 10 cents and 50 cents per million tokens. These are prices for different model releases, rather than a simple reduction for the same model.
Anthropic also reduced prices or costs across releases. Its Opus line moved from $5 input and $25 output per million tokens for Opus 5 in July to $4 and $20 for Opus 5.5. Anthropic also cut cache-read prices for some models. The company says Opus 5.5 costs 40% less than Opus 5 on typical workloads; that comparison is Anthropic’s claim, not an independent finding cited in the source material.
The customer examples come from The Information’s October 24 report, as described in public summaries, and follow-up reporting. Rippling CEO Parker Conrad said OpenAI’s share of the company’s coding work rose from 16% in May to 52%, while Anthropic’s share fell from 57% to 18%. The source material attributes the shift to cost, though it does not provide a breakdown of the contract terms behind it.
The reported deals differ from the headline model prices. The Information reported an OpenAI offer of a discount exceeding 30% on a contract worth more than $100 million annually for Replit, according to a person cited in the report. DataCamp CEO Jonathan Cornelissen said OpenAI’s offer could lower the company’s costs by 80% to 90%. That estimate compares expected total costs against DataCamp’s existing setup; it is not a stated 80% to 90% discount on OpenAI’s price list.
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
Lower Prices Reshape Vendor Choices
For businesses that buy large volumes of AI services, lower published rates and negotiated contract terms can change the cost of running products, coding tools and internal systems. The reported customer shifts suggest that price is one factor in provider selection, but companies also need to test whether a new model performs well on their own tasks before moving workloads.
The headline rate alone may not show what a customer will pay. Costs can depend on input and output usage, cache pricing, contract discounts, bundled seats and limits on included usage. A model with a lower per-token price may not produce the same total bill if it uses more tokens or performs differently. For customers, the immediate takeaway is to compare actual workload costs and contract terms, not just list prices.
Plan changes also affect users who are not negotiating enterprise contracts. The source reports that Anthropic introduced capped, expiring API credits in some Max and Team plans, while OpenAI reduced usage included with new $200 Pro subscriptions without changing that subscription’s price. Those changes can alter the amount of service available even when a monthly fee stays the same.
Products Worth Considering
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER)
As an affiliate, we earn on qualifying purchases.
A Fast Series of Price Changes
The reductions accumulated across several months rather than arriving in one announcement. In June, OpenAI previewed GPT-5.6 Sol and Luna at $5 and $30, and $1 and $6 per million input and output tokens, respectively. Anthropic launched Sonnet 5 at an introductory $2 and $10 rate that it later made permanent. In July, OpenAI cut Luna’s price, while Anthropic introduced Opus 5 at $5 and $25.
OpenAI later reduced Sol pricing and, in September, introduced GPT-6 Sol and Luna at lower rates. Anthropic introduced Opus 5.5 that month. On September 29, OpenAI cut the cache price for GPT-6.1 Sol. The source material says these figures were checked against vendor announcements and pricing pages, but model-to-model comparisons should be read carefully because newer releases may not match older models’ performance on a given task.
Cache costs can matter particularly for AI agents that repeatedly refer to prior conversation or task material. Data scientist Max Woolf estimated that cached tokens account for 96% to 99% of token spending in agentic coding. Anthropic says cache-related changes make some workloads less expensive, but those savings depend on the workload and the company’s own estimates. The price sequence is therefore not a single, uniform discount across every service or customer.
Deal Terms and Usage Limits
The public reporting does not give a complete account of the discounts, contract lengths, usage commitments or technical arrangements offered to each customer. The terms reportedly offered to Replit and a U.S. bank cannot be assumed to apply to other buyers. The 80% to 90% DataCamp estimate is also an expected saving, and the source material does not provide the underlying calculation or a final measured result.
It remains unclear how much of the customer movement reflects price, model quality, product features or other contract factors. Replit’s status is described differently across statements: Masad said the company was no longer working with Anthropic at an event, later called OpenAI its preferred partner and said Replit remained model agnostic; a person familiar with the partnership told TechCrunch it was still in place.
Published comparisons also do not establish that each newer model matches its predecessor on every task. Nor do lower token prices answer how much a specific customer will spend after allowances, caps, credits and seat requirements are applied. Anthropic’s Enterprise plan reportedly requires 20 seats, and OpenAI’s new $200 Pro subscriptions include less usage than before, according to the source material. The detailed effects of those terms for individual customers are not provided.
Watch Contracts and Real Usage
Businesses evaluating providers will need to compare their own workloads against current pricing and verify any negotiated rates, seat minimums, usage caps and credit-expiration rules in writing. They can also test competing models on the same tasks and track both output quality and total usage costs before shifting more work. That is especially relevant when a quoted saving depends on a customer’s existing setup rather than a standard public rate.
Further customer disclosures and changes to vendor pricing or plans may clarify whether the reported shifts are broad or limited to particular contracts. For now, the confirmed pattern is a series of lower prices across selected model releases, alongside reports of targeted offers and changing usage terms. The size and durability of savings for any one buyer remain dependent on its contract and actual use.
Key Questions
Did OpenAI cut its prices by 80% to 90%?
No general list-price cut of that size is established in the source material. DataCamp’s CEO described an expected 80% to 90% reduction in the company’s overall costs compared with its current setup.
How much did OpenAI’s Sol pricing fall?
The source material says OpenAI’s Sol line went from $5 input and $30 output per million tokens in June to $2 and $10 for GPT-6.1 Sol in September. The figures refer to different model releases.
Are companies switching from Anthropic to OpenAI?
Some reported examples show shifts in usage. Rippling CEO Parker Conrad said OpenAI’s share of coding work at the company rose to 52% from 16% in May, while Anthropic’s fell to 18% from 57%. The examples do not establish a market-wide pattern.
What should customers compare besides token prices?
Customers should check total workload costs, model performance on their tasks, cache rates, seat requirements, included usage, credit limits and expiration dates. Contract discounts can differ from published prices.
Source: ThorstenMeyerAI.com
