🔍 Read the full analysis: OpenAI And Anthropic Want Each Other’s Customers—Here’s What To Consider 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 is offering discounts to attract customers using Anthropic models, according to reporting by The Information, while both companies have reduced prices for some models. Reported customer savings are not all direct discounts: one CEO’s estimate of 80% to 90% lower costs refers to expected overall spending, while other reported offers include a discount of more than 30%.
Reported customer offers include a discount of more than 30% for Replit on a contract worth more than $100 million a year, according to a person cited by The Information. The Information also reported that a U.S. bank with more than $7 billion in assets received 500 ChatGPT seats at no extra charge, double its previous allocation. The source material does not provide the bank’s name or full contract terms.
DataCamp CEO Jonathan Cornelissen said OpenAI’s pricing would reduce the company’s AI costs by 80% to 90%. That is an estimate of expected savings against its existing setup, not a 90% discount off OpenAI’s price list. At Rippling, CEO Parker Conrad said OpenAI’s share of coding work rose from 16% in May to 52%, while Anthropic’s share fell from 57% to 18%. Conrad cited cost as the reason for the shift.
Published prices have also fallen, though comparisons may involve different model versions. OpenAI’s GPT-6.1 Sol is listed at $2 per million input tokens and $10 per million output tokens, compared with $5 and $30 for the earlier Sol model in June. Anthropic’s Opus price moved from $5 and $25 per million input and output tokens for Opus 5 to $4 and $20 for Opus 5.5. These figures describe token prices, not every customer’s total bill.
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
Price Cuts Reshape Vendor Choices
For companies buying AI, the combination of falling list prices and negotiated offers can change which provider is cheaper for a particular workload. A customer’s actual bill can depend on model performance, token use, cached input, support, contract commitments and included allowances—not just the published price per million tokens.
The reported customer shifts also show that vendor selection can change within a company as teams test models for specific tasks. Rippling’s figures describe the share of its coding work handled by each provider, not a broad measure of overall market share. Still, they illustrate why customers may reassess contracts when prices or model performance change.
Consumers and smaller teams face a related issue: a monthly subscription price can stay the same while usage limits or included features change. Comparing providers therefore means checking both the fee and what it buys.
Popular right now
De’Longhi Classic Matte Black Espresso Machine with Milk Frother
Samsung 42-Inch Class OLED S90F 4K Smart TV (2025)
Roborock Robot Vacuum and Mop with LiDAR Navigation and 10,000 Pa Suction
As an affiliate, we earn on qualifying purchases.
How Model Prices Have Shifted
Pricing changes accelerated over several months. Data scientist Max Woolf tracked cuts by both companies, and the source article says its review of vendor announcements and pricing pages matched the figures. OpenAI reduced prices for newer Sol and Luna models, while Anthropic made an introductory Sonnet price permanent and introduced later models at different rates.
Some cost claims concern cached tokens, which can matter in repeated exchanges with coding agents. Anthropic said its Fable 5.1 model could cost around 25% less on typical workloads and up to around 45% less on highly agentic workloads. Those are Anthropic’s estimates, not independently verified results for every customer. The source also cites an estimate that cached tokens account for 96% to 99% of spending in agentic coding; that figure is attributed to Woolf.
Price comparisons can also mix new models with their predecessors. Lower rates for a successor do not by themselves establish that it performs equally well on a customer’s tasks. The source article says OpenAI described GPT-6.1 Sol as offering “near-Astra intelligence for a fifth of the price”; customers would need to test that claim against their own work.
Deal Terms and Savings Vary
Full contract terms for the reported offers are not public in the supplied material. It is not clear what usage commitments, contract duration, model access or other conditions apply to the Replit and bank arrangements, or how the reported discounts compare across customers.
Replit’s relationship with Anthropic is also described differently by different accounts. Masad said at TechCrunch Disrupt that the company was no longer working with Anthropic, then told TechCrunch that Replit remained model agnostic and OpenAI was its preferred partner. A person familiar with the partnership told TechCrunch it was still in place. The source does not resolve the exact current scope of that arrangement.
Subscription terms are changing alongside token prices. The source says Anthropic added capped, expiring API credits to Max and Team plans, and that its Enterprise plan requires 20 seats. It also says OpenAI reduced usage included in new $200 Pro subscriptions without changing the price. The full limits and how they apply to each customer are not specified here.
Compare Offers Before Renewing
Companies evaluating a renewal or a new AI contract can compare their actual workload costs across vendors, including model quality, token use, cached-token rates, seat minimums, included usage and credit expiry. A short test using representative tasks can help separate a lower headline rate from a lower bill for the work a team actually performs.
Customers should also ask vendors to put discount duration, renewal rates, usage caps, overage charges and model access in writing. The reported offers suggest that negotiated terms may differ from published prices, but the available information does not show how widely those terms are being offered.
Further details may emerge as companies respond to the reported competition and clarify contract terms. For now, published prices and customer accounts show a fast-moving market, but they do not establish a single provider as the least expensive choice for every buyer.
Key Questions
Is OpenAI offering Anthropic customers a 90% discount?
Not on the evidence described here. DataCamp CEO Jonathan Cornelissen estimated that OpenAI’s pricing could lower his company’s overall AI costs by 80% to 90%. That is an expected saving against its existing setup, not a 90% reduction in OpenAI’s list price. The Information separately reported a discount of more than 30% for Replit.
How much have OpenAI and Anthropic’s model prices fallen?
The source reports OpenAI’s Sol rates moving from $5 per million input tokens and $30 per million output tokens in June to $2 and $10 for GPT-6.1 Sol. Anthropic’s Opus rates moved from $5 and $25 for Opus 5 to $4 and $20 for Opus 5.5. These are comparisons across model versions and may not predict a customer’s total costs.
Has Replit stopped using Anthropic?
The available accounts conflict on the scope of Replit’s current relationship. CEO Amjad Masad said at an event that Replit was no longer working with Anthropic, then told TechCrunch that the company remained model agnostic and OpenAI was its preferred partner. A person familiar with the partnership told TechCrunch it remained in place.
What should companies compare when choosing an AI provider?
Compare the cost and performance of models on the company’s own tasks, alongside contract discounts, usage caps, cached-token rates, seat minimums, credit expiry, overage charges and renewal terms. A low list price alone does not establish which service will cost less for a specific workload.
Source: ThorstenMeyerAI.com
