How To Read The Fine Print On OpenAI And Anthropic’s AI Plans

🔍 Read the full analysis: How To Read The Fine Print On OpenAI And Anthropic’s AI Plans on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets 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 on several AI models and are competing for business customers with reported discounts, but falling list prices do not tell buyers what a plan will cost in practice. A review published by ThorstenMeyerAI.com on October 26 says users also need to check usage allowances, credit expiration and seat minimums, which can change independently of a monthly price.

The review compared vendor pricing pages, retrieved October 26, with announced model prices. It says OpenAI’s Sol model moved from $5 input and $30 output per million tokens in June to $2 and $10 for GPT-6.1 Sol in September. Its Luna line moved from $1 and $6 to 10 cents and 50 cents per million tokens. These are prices for different model releases, not simply a price cut to an unchanged model.

Anthropic’s Opus line also became cheaper across releases: Opus 5 was listed at $5 input and $25 output per million tokens in July, while Opus 5.5 was listed at $4 and $20. Anthropic also cut cache-read prices for some models. The review says Anthropic attributes lower costs on typical workloads to those changes; the stated savings are vendor estimates, not a guarantee for every customer.

The review also summarizes customer deals reported by The Information on October 24. It says Replit was offered a discount of more than 30 percent on a contract reportedly worth more than $100 million annually. DataCamp CEO Jonathan Cornelissen described an expected 80 to 90 percent reduction in total costs under OpenAI’s offer. That figure is not a 90 percent reduction from OpenAI’s list price: it compares projected spending with DataCamp’s existing setup, according to the review’s account of the reporting.

Why AI Contract Terms Matter

Lower token prices can reduce bills, but the list price is only one part of total cost. Buyers may also pay for seats, usage above an allowance, or services and workloads that do not qualify for a particular discount. A low per-token figure does not establish how much a company will spend without its usage pattern and contract terms.

The review says OpenAI reduced usage included with new $200 Pro subscriptions while keeping the price unchanged. It also reports that Anthropic added capped, expiring API credits to its Max and Team plans, and that its Enterprise plan has a 20-seat minimum. Those terms can matter as much as a model’s headline rate, particularly for small organizations or individual users who do not need a large seat package.

For organizations, the reported competition may create room to seek better terms, but advertised or individually negotiated offers are not necessarily available to every buyer. Comparing providers on a representative workload, tracking actual usage, and checking renewal and overage rules gives procurement teams a more useful basis than comparing model prices alone.

How the Price Cuts Accumulated

The pricing moves came over several months in 2026. The review cites OpenAI price changes in July and August, Anthropic’s decision to make Sonnet 5’s introductory pricing permanent in August, and lower prices for later OpenAI and Anthropic models in September. It checked its timeline against vendor announcements and current pricing pages; the figures describe specific models and dates, rather than a single across-the-board rate.

That distinction matters because new models replaced older models in several comparisons. A lower price may accompany a new model with different performance, so buyers cannot assume it will produce equivalent results on their own tasks. The review says OpenAI described GPT-6.1 Sol as offering “near-Astra intelligence for a fifth of the price.” That is the company’s characterization; customers would need to test the model against their own requirements.

Some workloads also depend on cached input tokens, which may be priced differently from ordinary input. The review cites data scientist Max Woolf’s estimate that cached tokens account for 96 to 99 percent of token spending in agentic coding. That estimate is not a universal share for all AI use. It helps explain why cache prices can affect coding-agent bills even when a model’s headline input and output prices stay the same.

“I think ChatGPT is better now.”

— Replit CEO Amjad Masad, speaking at TechCrunch Disrupt on October 16, as quoted in the review

What Buyers Still Need to Verify

The reported discounts and savings are not directly comparable: the sources describe a negotiated contract discount, an expected reduction in total costs, and vendor list prices. Contract-specific terms remain undisclosed in the material provided, including how long reported offers last, which usage they cover, and what renewal rates apply.

The account of Replit’s relationship with Anthropic is also not fully settled in the cited reporting. Masad said Replit was no longer working with Anthropic, then described the company as model agnostic with OpenAI as its preferred partner; a person familiar with the partnership told TechCrunch it remained in place. The available material does not resolve the exact scope of that relationship.

It is also unclear how well each newer model performs on a given customer’s tasks, how much a buyer’s workload relies on discounted cached tokens, and whether plan allowances will change again. The review describes selected offers and terms; it does not establish that every customer can obtain the same pricing.

What to Check Before Renewal

Companies considering a renewal or new contract can ask providers to itemize costs by model, input and output tokens, cached usage, seats, and overages. They should also confirm credit caps and expiration dates, minimum seat counts, included usage, contract length, and the price after any promotional period. Testing candidate models on representative work can show whether a cheaper successor meets the same needs.

For individual subscribers, the immediate check is whether the plan’s included usage has changed even if its monthly fee has not. Buyers should compare the current terms in their own account or contract with the offer they are being asked to accept. No further pricing milestone is specified in the source material; the next developments to watch are vendor price-page changes and the terms customers receive in future negotiations.

Key Questions

Does the reported 80 to 90 percent figure mean OpenAI cut its list price by that much?

No. The review says DataCamp’s CEO described an expected reduction in total costs compared with the company’s current setup. That is different from a discount off OpenAI’s published prices.

Are the lower prices for the same AI models?

Not in every comparison. Several price changes involve new model releases replacing older ones, so customers should test performance on their own workloads rather than assume the models are interchangeable.

What plan terms should businesses check?

Check included usage, overage rates, seat minimums, credit limits and expiration dates, contract duration, and post-promotion pricing. Compare those terms with expected usage, not just the monthly fee or token rate.

Why do cached-token prices matter?

Some agentic workloads repeatedly reuse prior input, which may be billed at a separate cache-read rate. The review cites an estimate that cached tokens make up 96 to 99 percent of token spending in agentic coding; that estimate does not apply to every workload.

Source: ThorstenMeyerAI.com