Before Signing, Compare The Fine Print On OpenAI And Anthropic AI Plans

🔍 Read the full analysis: Before Signing, Compare The Fine Print On OpenAI And Anthropic AI Plans on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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

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 OpenAI is also offering customer-specific discounts as it seeks enterprise business, according to reporting cited by ThorstenMeyerAI.com. For buyers comparing plans, the reported price reductions are only part of the calculation: usage allowances, seat minimums and credit rules can change the total cost of a contract.

According to vendor pricing pages checked by ThorstenMeyerAI.com on October 26, OpenAI’s GPT-6.1 Sol costs $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 line fell from $1 and $6 per million tokens to $0.10 and $0.50. Anthropic’s Opus price moved from $5 and $25 per million tokens for Opus 5 to $4 and $20 for Opus 5.5. Anthropic also cut cache-read prices for some models.

The figures describe different models and pricing changes over time; they do not establish that a successor performs equally well on every customer’s tasks. The source says OpenAI described GPT-6.1 Sol as offering “near-Astra intelligence for a fifth of the price.” Buyers would need their own tests to judge performance for their workloads. Anthropic’s estimates that some models cost less on typical or agentic workloads are the company’s own claims, not independent measurements.

A report by The Information, published October 24, described OpenAI discounts aimed at attracting customers from Anthropic. The source reports that Replit received an offer of more than 30% off a contract valued at over $100 million a year, citing a person familiar with the report. DataCamp CEO Jonathan Cornelissen said he expected total AI costs to fall by 80% to 90% under OpenAI’s pricing. That is an expected saving against DataCamp’s existing setup, not a 90% reduction from OpenAI’s list price.

Other terms are changing alongside rates. 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 says OpenAI reduced the usage included with new $200 Pro subscriptions without changing the subscription price. Those details can affect buyers even when a headline monthly fee stays the same.

How Contract Terms Change the Cost

For a business buying AI services, the lowest advertised token rate may not yield the lowest bill. Costs depend on the amount and type of work, the model selected, how often an application reuses cached context, and whether the plan’s included usage covers demand. A contract with a discount can still be a poor fit if seat minimums or usage caps require the customer to pay for capacity it does not use.

The reported customer examples also show why the “80% to 90%” figure needs careful framing. It refers to one executive’s expected reduction in total costs under a specific offer, according to the source, rather than a standard discount available to all customers. Replit’s reported discount and a US bank’s reported seat arrangement are also individual deal terms, not general published rates.

For individuals, plan limits matter for a different reason: a subscription can retain its monthly price while the amount of usage included changes. Comparing plans therefore means checking current allowances and restrictions, not just the fee. Businesses may also reduce dependence on one vendor by testing more than one model and routing tasks to the option that meets their cost and performance needs.

Price Cuts Through September

The price changes followed a series of model launches and revisions between June and September 2026. The source cites data scientist Max Woolf’s October 25 account of the pricing sequence and says it checked the figures against vendor announcements and pricing pages retrieved the following day. OpenAI and Anthropic both changed prices during that period, while also releasing successor models.

Some changes affected cached input rather than headline input and output rates. Cached tokens can make up a large share of the cost in agentic coding, where a system repeatedly processes an ongoing conversation. Woolf estimates that cached tokens account for 96% to 99% of token spend in agentic coding. That estimate helps explain why cache pricing can matter to some workloads, but the share will depend on how a customer’s system operates.

The October 24 Information report added a separate development: negotiated discounts and customer movement. The source says Rippling CEO Parker Conrad reported that 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%. Those are figures attributed to Conrad, not a market-wide measure of adoption.

Limits of the Reported Savings

The source does not provide full contract documents or complete terms for the reported customer deals. The exact eligibility, duration, usage conditions and renewal provisions for OpenAI’s offers are not clear. Nor does the material establish that the reported discounts are available to other customers.

Several savings and performance figures are attributed estimates or company claims. DataCamp’s projected savings may reflect its own workload and current costs; Anthropic’s workload savings estimates are also its own. The source does not provide independent, comparable tests showing that each newer model delivers equivalent results for every use case. Replit’s relationship is described with differing statements: Masad said it was no longer working with Anthropic, later called OpenAI its preferred partner while saying Replit remained model agnostic, and a person familiar with the partnership told TechCrunch it was still in place.

Plan terms can change, and the source’s pricing-page check reflects October 26, 2026. The precise current usage allowances and credit conditions should be confirmed directly with each provider before a purchase or renewal.

Checks Before You Sign

Businesses weighing a new agreement can request a written schedule of rates, discounts, seat commitments, usage allowances, credit expiry dates and renewal terms. They can compare those terms against actual or representative workloads, including the cost of cached tokens where relevant, and test whether a less expensive successor model meets the required quality standards.

Customers should also ask whether negotiated prices apply throughout the contract or only for an introductory period, and what happens if usage exceeds a cap. The source does not identify a specific next pricing announcement or contract deadline from either company. For now, the practical next step is to verify current terms with the vendors and compare expected total spend rather than relying on headline rates alone.

Key Questions

Are OpenAI’s reported 80% to 90% savings available to every customer?

No. The figure is DataCamp CEO Jonathan Cornelissen’s estimate of his company’s expected total cost reduction under OpenAI’s pricing, as described in the source. It is not presented as a standard discount from OpenAI’s list price.

Which published model prices changed?

The source reports OpenAI Sol falling from $5/$30 per million input/output tokens in June to $2/$10 for GPT-6.1 Sol, and Luna falling from $1/$6 to $0.10/$0.50. It reports Anthropic Opus moving from $5/$25 to $4/$20. These are vendor prices cited by the source and refer to different model releases.

What plan details should buyers compare besides price?

Check included usage, seat minimums, API credit limits and expiry dates, renewal terms, and whether rates are temporary. The source says Anthropic’s Enterprise plan requires 20 seats and that OpenAI reduced usage included with new $200 Pro subscriptions without changing the price.

Do lower prices mean the newer models perform just as well?

Not necessarily for every task. The source says successor models replaced earlier ones and recommends testing them on a buyer’s own work. Vendor claims about capability or workload savings should be treated as claims unless independently verified for that use case.

Source: ThorstenMeyerAI.com