🔍 Read the full analysis: OpenAI And Anthropic Are Cutting AI Prices To Attract Rival Users 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 reduced prices for several models, and reporting describes OpenAI offering discounts to attract customers using rival AI services. Some businesses report changing how much work they send to each provider, but the widely cited 80% to 90% savings figure refers to one executive’s expected total cost reduction, not a discount off OpenAI’s list prices.
OpenAI and Anthropic have cut prices for several AI models in recent months, and OpenAI is also offering customer-specific deals to attract companies using rival services, according to reporting by The Information. The changes are reshaping costs for businesses that buy AI, but the headline figure of 80% to 90% savings describes one executive’s expected reduction in total costs—not a published discount off OpenAI’s standard prices.
Pricing pages reviewed by ThorstenMeyerAI.com on October 26 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. Its smaller Luna model moved from $1 and $6 to $0.10 and $0.50. These are prices for different model versions, so they do not show a like-for-like reduction for an unchanged model.
Anthropic also lowered prices across model launches. Opus 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. Anthropic also cut cache-read prices for some models. The company has said those changes can lower costs on typical workloads, with larger savings for highly agentic uses; those workload estimates are Anthropic’s own claims, not independent measurements.
The Information reported on October 24 that OpenAI is using discounts to pursue customers of Anthropic. Specific examples in accounts of that reporting include an offer of more than 30% off a Replit contract worth more than $100 million annually, and a U.S. bank receiving 500 ChatGPT seats at no additional charge. The source material does not provide the full terms of those arrangements. DataCamp CEO Jonathan Cornelissen said he expected OpenAI’s pricing to reduce his company’s costs by 80% to 90% against its current setup; that is an expected total-cost saving, not a stated list-price discount.
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
Lower Prices, More Contract Leverage
Falling prices can reduce the cost of using AI for coding, customer support and other business tasks. For companies with substantial usage, even a modest change in per-token rates can affect operating budgets. Customer-specific offers may also give buyers more leverage in negotiations, particularly if they can shift workloads between providers.
But list prices alone do not determine a bill. The amount of cached input, model choice, usage limits, seat counts and contract terms can all change the total. The source report also describes vendors adjusting allowances and credits, so a monthly subscription’s price may not tell customers how much capacity they receive. Buyers need to compare total cost for their actual workloads, rather than treating a lower posted rate as a complete measure of savings.
For users, the competitive pressure may bring more capable models into lower-cost plans, but the terms attached to those plans matter. A price that stays the same can still buy less if usage allowances change. For businesses, relying on a single provider may leave them exposed to future pricing or policy changes, while supporting multiple models can add testing and integration work.
Products Worth Considering
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
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
As an affiliate, we earn on qualifying purchases.
A Fast-Moving Price Timeline
The cuts followed a series of model releases and pricing updates from June through September 2026. Anthropic made Sonnet 5’s introductory price permanent in August, while OpenAI cut prices for Luna and later introduced lower prices for newer Sol and Luna versions. Anthropic’s September Fable 5.1 update reduced cache-read pricing, a cost component that can matter when AI agents repeatedly process conversation history.
Data scientist Max Woolf published a timeline of the changes on October 25. ThorstenMeyerAI.com said it checked the prices against vendor announcements and pricing pages retrieved October 26. Some advertised reductions compare a new model with its predecessor, rather than a price cut to an identical model. Whether a cheaper successor performs as well on a particular company’s work requires testing on that company’s own tasks.
The customer shift reports offer a separate measure of the competition. 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%. He cited cost as a reason. Replit CEO Amjad Masad said at TechCrunch Disrupt on October 16 that ChatGPT was better, then told TechCrunch that OpenAI was its preferred partner and that Replit remained model agnostic. TechCrunch also reported that a person familiar with the partnership said it was still in place.
“In May, Anthropic models handled 57 percent of the company’s coding work and OpenAI 16 percent; the reported shares later shifted to 18 percent for Anthropic and 52 percent for OpenAI.”
— Parker Conrad, Rippling CEO, as reported by The Information
Deal Terms and Savings Still Vary
The reported customer deals do not disclose complete contract terms, including how long discounts last, which models or usage levels they cover, or whether they include other services. The full details of the Information report are paywalled, and some figures are available through public summaries and secondary coverage. That makes it difficult to compare the reported offers directly with standard prices.
It is also unclear how much of the reported savings comes from lower token prices, changes in workload, negotiated credits or other contract provisions. The 80% to 90% figure is an executive’s estimate against his company’s current setup. It should not be read as a general savings rate available to all customers. Similarly, reported changes in model usage at Rippling and Replit do not establish that the same results would follow at other companies.
Finally, model versions differ. The price history compares successive releases, and vendor claims about performance or workload savings have not been independently established in the supplied material. Customers will need to test whether newer, cheaper models meet their requirements and account for any changes to subscription limits or expiring credits.
Watch Pricing and Usage Terms
Companies considering a switch can compare provider offers against their own usage data and test models on representative tasks before moving production workloads. That comparison should include input and output rates, cache pricing, included usage, seat minimums, credit expiration and the cost of operating with more than one provider.
OpenAI and Anthropic may make further changes as they compete for enterprise customers, but the source material does not identify a schedule for additional price announcements. Customers should check current vendor pricing and contract terms before making budget or purchasing decisions, since published rates and included allowances can change independently.
Key Questions
Are OpenAI’s prices 80% to 90% lower?
No. The 80% to 90% figure was DataCamp CEO Jonathan Cornelissen’s estimate of expected total savings against his company’s current setup. It was not a stated discount off OpenAI’s list prices.
What are the reported OpenAI price changes?
Pricing pages reviewed October 26 showed GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, compared with June prices of $5 and $30 for an earlier Sol version. Luna moved from $1 and $6 to $0.10 and $0.50 across the versions cited. These comparisons involve different model releases.
Has Anthropic also reduced prices?
Yes. The source material lists Opus 5 at $5 per million input tokens and $25 per million output tokens, and Opus 5.5 at $4 and $20. Anthropic also lowered cache-read prices for some models and has published its own estimates of resulting workload savings.
Are companies moving from Anthropic to OpenAI?
Some reported customer activity points to shifts, but it does not establish a broad migration. Rippling CEO Parker Conrad reported a change in the share of coding work handled by each provider. Replit described OpenAI as its preferred partner while saying it remained model agnostic; TechCrunch reported that the partnership with Anthropic was still in place.
What should AI buyers compare before switching?
Compare the full cost of actual workloads, including token and cache rates, usage limits, seats, credits and contract duration. Test newer models on representative tasks, because a lower price for a successor model does not by itself establish equivalent performance.
Source: ThorstenMeyerAI.com
