AI Subscription Shopping? Read The Fine Print On OpenAI And Anthropic

🔍 Read the full analysis: AI Subscription Shopping? Read The Fine Print On OpenAI And Anthropic on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the little things that make your day 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 in recent months, while new reporting describes OpenAI offering contract discounts to attract enterprise customers. The price changes could lower bills for companies and developers, but reported savings vary by customer, and subscription limits and contract conditions can change the actual cost.

According to vendor pricing pages checked by ThorstenMeyerAI.com on October 26, OpenAI’s Sol model pricing 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 moved from $1 and $6 to 10 cents and 50 cents per million input and output tokens. Those figures compare different model versions, rather than showing the same model’s price alone falling.

Anthropic also lowered some prices. 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. The source also reports lower cache-read costs for some models. Anthropic has said these changes can reduce costs on typical or agentic workloads, but those savings are its estimates and depend on how customers use the models.

Separately, The Information reported on October 24 that OpenAI has offered discounts to attract customers using Anthropic models. The reported examples are not all equivalent: one person cited by The Information said Replit received an offer of more than 30% off a contract worth more than $100 million a year. DataCamp CEO Jonathan Cornelissen said he expected OpenAI’s pricing to reduce his company’s costs by 80% to 90%. That is an expected reduction in total costs against the company’s existing setup, not a 90% reduction from OpenAI’s list price.

The Contract Can Outweigh the List Price

For businesses buying AI at scale, falling per-token prices can make it worthwhile to revisit contracts and test alternatives. But the headline rate is only one part of the bill. Costs depend on the models used, the volume and type of work, and whether cached input, seats or usage allowances are included. A lower price per token does not automatically mean the same task will cost less if a new model uses tokens differently or performs differently on the buyer’s work.

Subscription terms also affect the comparison. The source says Anthropic added capped, expiring API credits to its Max and Team plans, and that its Enterprise plan has a 20-seat minimum. It also reports that OpenAI reduced the usage included with new $200 Pro subscriptions while keeping the monthly price unchanged. These examples show why customers should compare included usage, credit expiry, minimum seats and overage rules alongside the monthly fee.

Competition may give some large buyers leverage, but the reported offers are specific deals, not a published discount schedule available to everyone. Smaller firms and individual subscribers may see different pricing and limits. For buyers, the practical question is the cost of completing their own workload under the full contract, not which provider has the lowest advertised number.

How Model Prices Changed This Year

The source’s timeline describes a sequence of model launches and price revisions between June and September 2026. It cites data scientist Max Woolf’s October 25 pricing review and says its figures were checked against vendor announcements and pricing pages retrieved on October 26. The sequence includes introductory pricing for Anthropic’s Sonnet 5, later made permanent, and successive OpenAI price changes for its Sol and Luna models.

Some reductions involve new models replacing older versions, rather than a price cut for an unchanged product. Performance can differ between versions, so companies need to test models against their own tasks before treating a lower listed rate as a like-for-like saving. Cache pricing also matters in repeated-turn workflows such as coding agents. Woolf estimates cached tokens account for 96% to 99% of token spending in agentic coding; that is his estimate, not a universal figure for all AI use.

Customer movement is also part of the reported shift. 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%. The figures describe Rippling’s coding work, not the overall market. Replit’s relationship with OpenAI has also drawn attention: CEO Amjad Masad called OpenAI its preferred partner while later telling TechCrunch that Replit remained model agnostic; a person familiar with the partnership told TechCrunch it was still in place.

What Buyers Still Need to Verify

The available reporting does not establish that the deals cited are available to other customers, or disclose every contractual condition, including duration, usage thresholds and renewal terms. The Information’s article is paywalled, and the source account says it relied on its public summary and additional reports when describing some of the offers. The reported savings should be read as customer-specific claims, not independently verified outcomes for the wider market.

It is also unclear how each new model performs on individual customers’ tasks compared with the model it replaces. Published token prices do not show the full cost of a workflow, and vendor estimates of savings may not apply to every use case. Subscription allowances and API-credit rules can also change; buyers should confirm current terms directly with the provider before making a decision.

Compare Workloads Before Renewing

Companies considering a switch or contract renewal can test more than one vendor on representative tasks, then compare output quality, usage and total cost under the proposed terms. They should ask providers to spell out seat minimums, included usage, credit expiry, overages and any limits on discounts. The source recommends keeping work routable across vendors where practical, which can give buyers more flexibility if prices or terms change.

For individual subscribers, the next step is to check what the plan includes now rather than relying on its monthly price alone. The source does not identify a specific upcoming price announcement or policy deadline from either company. The next developments to watch are vendor pricing-page changes, revised plan allowances and further customer reports about how negotiated offers perform in practice.

Key Questions

Did OpenAI give customers a 90% discount?

Not as a general list-price discount. The 80% to 90% figure refers to DataCamp CEO Jonathan Cornelissen’s estimate of the company’s expected total cost reduction against its existing setup. Other reported offers include a discount of more than 30% for Replit.

Which prices changed?

The source reports lower prices for several model versions from both companies. OpenAI’s Sol and Luna prices fell between June and September, while Anthropic’s Opus 5.5 was listed below Opus 5. These comparisons involve different model versions and may not represent identical performance or workloads.

Why do subscription terms matter if token prices are lower?

A plan’s total value also depends on its usage allowance, seat minimums, credit limits and expiry rules. The source reports changes to credits and subscription usage, so a lower advertised model price may not mean a lower overall bill for every customer.

Are customers moving from Anthropic to OpenAI?

Some reported examples show a shift in use. Rippling CEO Parker Conrad said OpenAI’s share of the company’s coding work rose to 52% from 16% since May, while Anthropic’s fell to 18% from 57%. Those figures describe Rippling, not the market as a whole.

Source: ThorstenMeyerAI.com