🔍 Read the full analysis: Before You Pick An AI Service, Compare OpenAI And Anthropic’s Pricing on ThorstenMeyerAI.com
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
TL;DR
OpenAI and Anthropic have cut prices for several AI models since June 2026, and reporting describes OpenAI offering customer-specific discounts as it competes for business. Those deals are not the same as across-the-board list-price cuts: customers also need to compare usage limits, seat minimums, credits and model performance.
OpenAI and Anthropic have lowered prices for several AI models since June, while OpenAI is also offering some business customers negotiated discounts, according to vendor pricing pages and reporting by The Information. The changes matter to companies choosing AI services, but headline token prices alone do not show what customers will pay: usage allowances, contract terms and the performance of replacement models also affect total cost.
Pricing figures compiled by data scientist Max Woolf show OpenAI’s Sol model moving 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. OpenAI’s Luna line fell from $1 and $6 to $0.10 and $0.50 per million tokens. Anthropic’s Opus pricing moved from $5 and $25 for Opus 5 in July to $4 and $20 for Opus 5.5. Anthropic also reduced cache-read prices for some models.
These comparisons cover different model versions, not necessarily identical models becoming cheaper. Anthropic has said its Fable 5.1 model can cost about 25% less on typical workloads and up to about 45% less on highly agentic workloads; those are company estimates. Woolf’s analysis says cached tokens can account for 96% to 99% of token spending in agentic coding, making cache prices relevant to some customers. Actual savings depend on how a service is used.
The Information reported on October 24 that OpenAI was using discounts to compete for large customers. The reported examples include a discount of more than 30% for Replit on a contract worth more than $100 million a year, according to a person cited in the report, and a deal providing a US bank with 500 ChatGPT seats for no extra charge. These are reported contract terms, not general offers available to all customers.
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 AI Contracts Need Fresh Comparisons
The price changes give buyers a reason to revisit contracts and test alternatives, but they do not establish that one provider is cheaper for every workload. Businesses may pay for API tokens, subscriptions, seats, or negotiated enterprise packages, and each structure can produce a different effective cost. A lower per-token rate may matter little if a plan’s usage cap is restrictive; a discounted seat may be valuable only if employees use it.
Reported customer moves suggest price and model preferences can affect where companies direct work. 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 one company’s reported usage, not an industry-wide shift, and do not by themselves establish that price was the sole cause.
For buyers, the practical comparison is the total cost for a defined workload, alongside quality, reliability, data terms and the ability to switch providers. Testing models on the company’s own tasks can help establish whether a lower-priced successor performs well enough before a broad change is made.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
Price Cuts and Plan Terms
The reported reductions came through several steps rather than one across-the-board price announcement. OpenAI lowered Luna pricing in July, made a temporary Sol price cut in August, and introduced lower-priced GPT-6 models in September. Anthropic introduced Sonnet 5 at introductory pricing in June and made that price permanent in August, before releasing Opus 5.5 in September. Woolf published his price timeline on October 25; the source article says its figures were checked against vendor announcements and pricing pages retrieved October 26.
Pricing is only one part of the comparison. Anthropic has added capped, expiring API credits to its Max and Team plans, according to the source material, while its Enterprise plan requires 20 seats. OpenAI reduced the usage included with new $200 Pro subscriptions without changing the monthly price, the source says. Such terms can change the value of a plan even when its advertised fee stays the same.
Some striking savings figures need careful attribution. DataCamp CEO Jonathan Cornelissen described expected total cost savings of 80% to 90% from OpenAI’s offer, according to reporting cited by the source. That is an estimate against DataCamp’s existing setup, not a 90% reduction from OpenAI’s public list price.
Which Savings Apply to Buyers
The reported deals do not disclose all contract details, including the eligibility requirements, duration, usage conditions or renewal pricing. The source material does not specify the terms behind every customer offer, so other companies cannot assume they can secure the same rates.
It is also unclear how broadly customers are shifting work between providers. Rippling’s reported coding-work shares are one company’s figures, and the available information does not provide a comparable survey across businesses. Replit’s public remarks and later clarification describe a preferred partner while retaining a model-agnostic position; the exact scope of its current partnership is not fully established by the source material.
Finally, lower prices for successor models do not prove they deliver equivalent results for every task. Model quality, token use and cache behavior vary by workload. Buyers need their own tests to estimate whether the published rates translate into savings without reducing performance or changing other costs.
Recheck Usage and Renewal Terms
Companies comparing vendors should price a representative set of tasks across more than one model, then review the contract’s seat minimums, usage allowances, credits, expiration dates and renewal terms. For existing customers, the reported market competition may provide an opening to ask providers for revised pricing or clearer usage terms, though no particular discount is guaranteed.
Users on subscription plans should check whether their allowance has changed, even if the monthly fee has not. The next useful evidence will be updated vendor pricing and plan documentation, alongside disclosed contract terms and customer reports that show whether savings persist after initial offers or model changes.
Key Questions
Did OpenAI cut its prices by 90%?
No general 90% list-price cut is established here. The figure refers to DataCamp CEO Jonathan Cornelissen’s expected total savings against the company’s current setup, as reported in the source material. OpenAI’s published token-price changes are separate.
Which OpenAI prices changed?
The source’s comparison says 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 $0.10 and $0.50. The figures compare different model versions.
Are Anthropic’s new prices lower for every customer?
Published model prices changed, but customer costs depend on usage and plan terms. The source also reports capped, expiring credits on some plans and a 20-seat minimum for Enterprise. Individual contract terms may differ.
How should a business compare AI services?
Test providers on representative company tasks and compare total expected costs, not only token rates. Include subscription allowances, seats, cache pricing, credit limits, contract duration and renewal terms.
Source: ThorstenMeyerAI.com
