🔍 Read the full analysis: OpenAI And Anthropic Are Competing On Price—What Customers Should Know on ThorstenMeyerAI.com
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
TL;DR
OpenAI and Anthropic have lowered prices for several AI models since June, and reporting describes OpenAI offering additional discounts to win enterprise customers. Those reported savings are not all list-price cuts: customers should compare contract terms, usage limits, credits and performance, as well as token prices.
OpenAI and Anthropic have cut prices for several AI models since June, while OpenAI is also offering discounted enterprise deals to compete for customers, according to reporting by The Information. The changes could lower bills for businesses using AI, but the headline savings do not tell the full story: contract discounts, usage allowances, credits and model performance all affect the actual cost.
Pricing pages and announcements reviewed by ThorstenMeyerAI.com show OpenAI’s Sol model falling 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 Luna model moved from $1 and $6 to 10 cents and 50 cents per million input and output tokens. These are prices for successive models, not necessarily reductions to the same model while its capabilities stayed constant.
Anthropic also reduced prices across its model lineup. Opus went 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 the price of cached input for that model by 60 percent, according to the source. Cache pricing can matter for repeated prompts and agentic workflows, where a system may process prior conversation or task context again.
The enterprise offers reported by The Information are separate from those public list prices. The Information reported a discount of more than 30 percent for Replit on a contract valued at over $100 million annually, and a deal giving a U.S. bank 500 ChatGPT seats for less than $7 per seat. DataCamp CEO Jonathan Cornelissen said he expected OpenAI’s offer to reduce his company’s costs by 80 to 90 percent. That figure describes expected savings against DataCamp’s existing setup, not a 90 percent reduction from OpenAI’s published prices.
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
Why Enterprise AI Bills Are Shifting
Falling public prices and negotiated deals give buyers more leverage when renewing contracts or choosing models. The reported customer changes suggest the competition is affecting how companies assign work: Rippling CEO Parker Conrad said OpenAI’s share of his company’s coding work rose from 16 percent in May to 52 percent, while Anthropic’s share fell from 57 percent to 18 percent. He attributed the shift to cost.
For customers, however, a lower token price does not automatically mean a lower bill for equivalent work. Teams need to test whether a newer model performs adequately on their tasks, estimate the volume and mix of input and output tokens, and account for cache charges, service limits and contract terms. A discount can be valuable, but only if it applies to the relevant usage and does not come with restrictions that offset the savings.
The competition also makes vendor concentration a practical concern. Routing different tasks to more than one provider may give a buyer flexibility and negotiating power, although it can add engineering and oversight work. The best comparison is the cost and performance of completing a defined workload—not a single headline price.
Products Worth Considering
Withings Sleep Under-Mattress Sleep Tracking Pad
Withings Body Scan Smart Scale with Segmental Body Composition and ECG, Black
LEVOIT Vital 100S-P Smart Air Purifier with Washable Pre-Filter
As an affiliate, we earn on qualifying purchases.
How Prices And Plans Changed
The reductions followed several changes between June and September 2026. Anthropic introduced Sonnet 5 at an introductory price of $2 per million input tokens and $10 per million output tokens, then made that price permanent, according to the source material. OpenAI cut Luna pricing in July and reduced Sol pricing in August before introducing lower prices for the GPT-6 models in September.
Not every cost change appeared in the main input and output rates. Anthropic said its Fable 5.1 model’s lower cache-read price could reduce costs by around 25 percent on typical workloads and up to around 45 percent on highly agentic ones. Those estimates are Anthropic’s claims, not independent measurements. The source also reports Anthropic describing Opus 5.5 as costing 40 percent less than Opus 5 on typical workloads.
Plan details can move alongside model prices. According to ThorstenMeyerAI.com, Anthropic added capped, expiring API credits to its Max and Team plans, while its Enterprise plan requires a 20-seat minimum. OpenAI reduced the usage included with new $200-per-month Pro subscriptions without changing the monthly price, the source reports. Those terms can affect smaller customers and heavy users even when the listed subscription fee stays the same.
“OpenAI’s share of the company’s coding work rose from 16 percent in May to 52 percent, while Anthropic’s fell from 57 percent to 18 percent.”
— Parker Conrad, Rippling CEO, as reported by ThorstenMeyerAI.com
What Buyers Still Need To Verify
The reported commercial terms are not a full account of the contracts. Public reporting does not establish every customer’s discount conditions, duration, usage thresholds or other fees, and the reported bank offer is not accompanied here by details of the bank’s contract. The Information article is paywalled; ThorstenMeyerAI.com said it reviewed the public summary and secondary coverage when discussing its reporting.
It is also unclear how broadly the reported offers are available. A deal negotiated for a large customer may not be accessible to a smaller organization, and a large expected saving at one company cannot be assumed for another. Model comparisons also depend on the customer’s tasks: the source notes that successor models may differ in capability, so buyers should validate performance on their own workloads before switching.
Plan limits and credit rules may change over time. The available material identifies some changes but does not provide a complete, current comparison of every vendor plan. Customers should confirm the terms directly with providers and check whether stated allowances, credits and prices apply to their intended usage.
What To Check Before Renewing
Organizations considering a renewal or a new AI contract can use the price changes as a reason to request updated terms from each provider. They should compare total cost for representative workloads, including input and output usage, cached tokens, seat minimums, included allowances, overages and credit expiration. Where a vendor quotes a percentage saving, buyers should ask what baseline and time period the calculation uses.
Teams can also test more than one model on the same tasks and track both cost and results. That helps separate a lower price per token from a genuine reduction in the cost of completing work. Maintaining a second provider for suitable tasks may preserve flexibility, provided the additional integration and management costs are included.
The next developments to watch are further list-price or plan changes, and whether more companies disclose contract terms or shifts in their model usage. Until those details are public, the reported deals show that competition is intensifying but do not establish a standard discount available to all buyers.
Key Questions
Did OpenAI cut prices by 90 percent?
No general 90 percent list-price cut is established here. DataCamp’s CEO described expected total savings of 80 to 90 percent against the company’s existing setup. The Information also reported separate customer deals, including a discount of more than 30 percent for Replit.
Which OpenAI prices fell?
According to pricing information reviewed by ThorstenMeyerAI.com, Sol moved from $5/$30 to $2/$10 per million input/output tokens between June and September, while Luna moved from $1/$6 to 10 cents/50 cents. These comparisons involve successive models.
Did Anthropic lower its prices too?
Yes. The source reports that Opus pricing moved from $5/$25 to $4/$20 per million input/output tokens between Opus 5 and Opus 5.5. Anthropic also reduced cache-read pricing for Opus 5.5, according to the source.
What should businesses compare besides token prices?
Compare the cost of completing real tasks, including usage allowances, cache rates, seat minimums, expiring credits, overage charges and contract terms. Test model performance on your own work before deciding that a lower-priced successor is an equivalent replacement.
Source: ThorstenMeyerAI.com
