🔍 Read the full analysis: AI Prices Are Falling As OpenAI And Anthropic Battle For Customers on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
OpenAI and Anthropic have cut prices for several AI models as they compete for business customers, and reporting describes OpenAI offering discounts to some large accounts. The widely cited 80% to 90% savings figure is a customer’s expected total cost reduction, not a general discount from OpenAI’s list prices; contract terms and included usage also matter.
OpenAI and Anthropic have lowered prices for several AI models in recent months, while OpenAI is also offering discounts to some business customers as it competes with Anthropic for accounts. The changes are reducing costs for some buyers, but reported savings vary by customer and workload, and subscription limits and contract terms are changing alongside list prices.
Vendor pricing pages reviewed on October 26 show OpenAI’s GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, compared with $5 and $30 for the earlier Sol model in June. Its Luna model moved from $1 and $6 per million tokens to $0.10 and $0.50. The figures compare different model versions, rather than showing a like-for-like reduction for an unchanged model.
Anthropic also reduced prices across its model line. Opus pricing moved 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 cut the price of cached input tokens for some models. The company says those changes reduce costs on typical workloads; the estimates are its own and will depend on how customers use the models.
Separate reporting by The Information, published October 24, describes OpenAI pursuing Anthropic customers with individual offers. The reported examples include a discount of more than 30% for Replit and seats priced below $7 for a US bank. These are reported account-specific deals, not published terms available to every customer. DataCamp CEO Jonathan Cornelissen described expected savings of 80% to 90% from OpenAI’s offer, but that figure refers to the company’s expected total costs against its existing setup, not a 90% reduction in OpenAI’s 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
Lower Prices, More Contract Tradeoffs
Lower per-token prices can reduce the cost of deploying AI for coding, customer support and other work that consumes large volumes of model output. For buyers, the competition also creates an opening to renegotiate existing contracts or compare offers from multiple providers. The reported customer moves suggest that cost is affecting which models companies assign to work, though it is not the only possible factor.
List prices alone do not capture the full cost. Coding agents can repeatedly process prior conversation context, making cached-token pricing and usage patterns relevant. A cheaper model may also perform differently on a company’s specific tasks. Businesses need to compare total spending, performance and included usage, rather than treating a lower per-token rate as proof of equivalent results or savings.
Plan terms can offset headline price reductions. Anthropic has added capped, expiring API credits to its Max and Team plans, and its Enterprise plan requires a minimum of 20 seats, according to the source material. OpenAI has reduced the usage included with new $200 Pro subscriptions without changing that monthly price. Buyers and individual subscribers may therefore pay the same fee while receiving a different allowance.
Popular right now
Shark Rocket Pet Plus Corded Stick Vacuum, Bordeaux, HV322
De’Longhi Classic Matte Black Espresso Machine with Milk Frother
Samsung 42-Inch Class OLED S90F 4K Smart TV (2025)
As an affiliate, we earn on qualifying purchases.
How the Price Cuts Accumulated
The price changes arrived over several months in 2026. OpenAI reduced pricing for its Luna model in July and lowered Sol pricing in August before September releases brought lower rates for newer Sol and Luna models. Anthropic introduced Sonnet 5 at an introductory price in June, made that price permanent in August and released lower-priced Opus 5.5 in September. Its Fable 5.1 release also reduced cache-read costs, according to the pricing timeline cited by the source.
Data scientist Max Woolf published a price timeline on October 25. The source says its author checked the figures against vendor announcements and pricing pages retrieved October 26. That timeline covers successive model releases and pricing adjustments; it should not be read as a direct comparison of identical models over time. The source also reports Anthropic’s estimate that cached tokens account for most token spending in agentic coding workloads, but that estimate is not an independent measurement of every customer’s usage.
Customer accounts provide another view of the contest. 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%. Replit CEO Amjad Masad said at TechCrunch Disrupt that ChatGPT was better for the company, then told TechCrunch that OpenAI was its preferred partner and Replit remained model agnostic. The source also notes that a person familiar with the partnership told TechCrunch the relationship remained in place.
“I think ChatGPT is better now.”
— Replit CEO Amjad Masad, speaking at TechCrunch Disrupt
Deal Terms and Savings Still Vary
The full terms of the reported OpenAI offers are not public in the supplied material, including which services, usage volumes and contract periods the discounts cover. The reported examples concern particular customers and cannot be assumed to represent standard pricing. The reported 80% to 90% DataCamp savings estimate is a comparison with that company’s existing setup, and the source does not provide enough detail to independently calculate it.
It is also unclear how widely the customer shifts described in the reports extend beyond the named companies. Model usage shares at Rippling reflect one company’s coding work, while the Replit comments and partnership reports do not establish that the company has completely ended its relationship with Anthropic. Finally, model successors and workload-dependent cache savings make direct price comparisons difficult: customers must establish whether a lower-cost model meets their own quality and reliability needs.
Compare Offers Against Real Usage
Companies evaluating the changes can compare vendor proposals against their actual usage, including input, output and cached-token volumes, seat counts, credit expiration and usage caps. Testing models on representative tasks can show whether a lower-priced successor delivers acceptable results before a buyer shifts substantial work or signs a longer contract. Using more than one provider may also give businesses options when prices or terms change.
Customers will need to watch for further revisions to model pricing and subscription allowances. The source material does not identify a scheduled price announcement or a confirmed next milestone from either company. For now, the clearest indicators are vendors’ updated pricing pages, published plan terms and the final conditions customers negotiate directly.
Key Questions
Are AI model prices falling?
Several listed prices have fallen at OpenAI and Anthropic, based on vendor pricing information cited in the source. Many comparisons involve newer models replacing older ones, so buyers should not assume that every model became cheaper while staying otherwise unchanged.
Does the 80% to 90% figure mean OpenAI cut its prices by that much?
No. The figure is DataCamp CEO Jonathan Cornelissen’s estimate of the company’s expected total savings against its existing setup. It is not a general 80% to 90% discount from OpenAI’s published rates.
What business discounts have been reported?
The source cites a reported discount of more than 30% for Replit and seats below $7 for a US bank. These are specific reported deals; their full terms are not public in the supplied material.
Why do cached-token prices matter?
Some AI workflows repeatedly process earlier conversation context. When that context qualifies for cached-token pricing, its rate can affect the total bill. The savings depend on a customer’s workload and the provider’s cache rules.
What should customers compare besides the model price?
Compare performance on representative tasks, total token usage, cached-token rates, seat minimums, included usage, credit caps and expiration dates. Subscription allowances can change even when the monthly fee stays the same.
Source: ThorstenMeyerAI.com
