The AI Price War: OpenAI And Anthropic’s Push To Win New Subscribers

🔍 Read the full analysis: The AI Price War: OpenAI And Anthropic’s Push To Win New Subscribers on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get audio and creator gear delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

OpenAI and Anthropic have cut prices for several AI models since June, and OpenAI is also offering discounts to attract business customers, according to reporting summarized by ThorstenMeyerAI.com on October 26. The changes are lowering some customers’ costs and shifting where companies send AI work, but headline savings figures do not all refer to the same thing: they range from published token prices to negotiated contract terms and one customer’s projected total savings.

According to the vendors’ pricing pages, OpenAI’s Sol model family 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. Its smaller Luna model went from $1 and $6 to 10 cents and 50 cents per million input and output tokens. These are prices for successive models, so the figures do not show the same model becoming cheaper while remaining unchanged in capability.

Anthropic also reduced prices across its model lineup. Its Opus pricing 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 cache-read prices for some models. The company says those changes reduce costs on typical workloads, but the savings depend on how a customer uses the models.

The customer figures in the October 24 report by The Information describe different arrangements, not a universal discount. Rippling CEO Parker Conrad said OpenAI’s share of his company’s coding work rose from 16% in May to 52%, while Anthropic’s fell from 57% to 18%, citing cost. The Information also reported an OpenAI discount of more than 30% for Replit on a contract worth over $100 million a year. DataCamp CEO Jonathan Cornelissen described expected savings of 80% to 90% against his company’s existing setup; that is a projected total-cost reduction, not a 90% cut to OpenAI’s list price.

Lower Prices, Different Customer Terms

The cuts give businesses more room to compare providers and negotiate contracts, particularly when AI use accounts for substantial software spending. But a lower token rate does not automatically mean a lower bill: usage volume, cached tokens, model performance and contract terms all affect total cost. A company may also need to test whether a newer, cheaper model performs adequately for its own work before moving workloads.

For individual subscribers, the monthly price is only part of the deal. The source reports that OpenAI reduced the usage included with new $200 Pro subscriptions without changing the monthly fee. Anthropic added capped, expiring API credits to its Max and Team plans. These changes make allowances and limits worth checking alongside the advertised price, especially for customers whose usage is near a plan’s cap.

Competition can benefit buyers through lower rates and added capacity, but switching providers can bring technical and commercial costs of its own. Companies need to compare what they will pay for their actual workloads—not just a price per million tokens—and account for the possibility that introductory prices or negotiated discounts carry conditions.

How the 2026 Price Cuts Unfolded

The reported cuts arrived over several months. OpenAI reduced Luna pricing in July, cut Sol pricing temporarily in August, and introduced lower prices for GPT-6 models in September. Anthropic made Sonnet 5’s introductory pricing permanent in August and launched Opus 5.5 at lower listed rates than Opus 5 in September. The sequence includes new models, promotional prices and changes to specific token categories, which are not directly interchangeable.

One less visible cost factor is cached input. Coding agents may repeatedly process earlier conversation or task information, and cached tokens can be priced differently from ordinary input. Data scientist Max Woolf estimated that cached tokens account for 96% to 99% of token spending in agentic coding. Anthropic has said its Fable 5.1 model can cost around 25% less on typical workloads and up to around 45% less on highly agentic workloads; those are the company’s estimates, not independent measurements.

The October 24 Information report added a separate development: reported offers aimed at business customers already using competing models. Its account, as described in the supplied source, includes customer statements and deal details attributed to people familiar with negotiations. Those reports should not be confused with public list prices, which are available to all customers under published pricing schedules.

Discount Sizes and Plan Limits

The full terms of the customer offers described in The Information’s report are not available in the supplied material. The account is paywalled, and the reported discount of more than 30% for Replit is attributed to a person cited by the publication. The source does not provide a complete contract or explain all conditions attached to the offer.

The 80% to 90% DataCamp figure is an expected reduction in total costs, but the baseline calculation and the workloads included are not detailed here. It is not possible to apply that estimate to other customers. Nor do the published token prices establish how much a typical business will save after factoring in usage, caching, model quality and plan allowances.

Replit’s relationship with Anthropic is also described differently across accounts. Masad said 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 was still in place. The exact status and scope of that relationship are not clear from the available reporting.

What Buyers Should Check Next

Businesses considering a switch can compare providers using representative tasks and recent bills, including separate calculations for ordinary input, output and cached tokens. They can also ask vendors to spell out discount duration, minimum commitments, usage caps and renewal pricing before signing. Routing work across more than one provider may offer flexibility, though it adds operational and testing requirements.

Subscribers should review current plan terms rather than assume that a monthly fee guarantees the same usage allowance over time. The source says Anthropic’s Enterprise plan requires 20 seats, a threshold that could put it beyond the reach of many small advisory firms; it also reports changes to OpenAI’s usage included with new Pro subscriptions. Customers should confirm the applicable terms with each vendor.

The next useful signals will be more detailed disclosure about business contracts, further changes to published pricing and evidence from customers comparing costs on real workloads. Until those details are available, the reported price war shows stronger competition but does not establish a single savings rate that applies across the market.

Key Questions

Did OpenAI cut its prices by 90%?

No, not according to the supplied reporting. DataCamp’s CEO described expected total cost savings of 80% to 90% against the company’s existing setup. The figure is not a 90% reduction in OpenAI’s published list prices.

How much did OpenAI’s listed model prices change?

According to the vendor pricing pages cited in the source, Sol moved from $5 input and $30 output per million 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 prices apply to successive models, not the same model held constant.

Are companies moving coding work from Anthropic to OpenAI?

Some reported customer data points that way. 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 concern Rippling and should not be read as industry-wide usage data.

What should customers compare besides token prices?

Customers should check usage allowances, cache pricing, seat minimums, credit expiration, discount conditions and renewal terms. They should also test model performance on their own tasks, since lower prices for a successor model do not by themselves establish that it will work equally well for every use.

Source: ThorstenMeyerAI.com