Why OpenAI And Anthropic Are Cutting Prices To Win AI Customers

Read the full analysis: Why OpenAI And Anthropic Are Cutting Prices To Win AI Customers on ThorstenMeyerAI.com

OpenAI and Anthropic have cut prices for several AI models since June, and reporting published this weekend says OpenAI is also offering customer-specific discounts to attract business from Anthropic. The changes could reduce costs for companies using AI, but headline prices do not capture the full bill: usage limits, credits, seat minimums and model performance also affect what customers pay.

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. OpenAI’s Luna line fell from $1 and $6 per million tokens to $0.10 and $0.50. These prices are for different model versions, not simply lower rates for unchanged models.

Anthropic has also reduced prices. Its Opus model 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. The company also cut cache-read pricing for some models. Anthropic says Opus 5.5 costs 40% less than Opus 5 on typical workloads; that is a vendor estimate, not an independently established saving for every customer.

The Information reported on October 24 that OpenAI is using discounts to pursue customers associated with Anthropic. The source article’s reported examples include a discount of more than 30% for Replit on a contract valued above $100 million a year, and a US bank receiving sub-$7 seats, according to ThorstenMeyerAI.com’s account of the reporting. The account also reports a bank’s ChatGPT seat count rose to 500 without an added charge. The exact terms were not disclosed publicly.

Lower Prices, Closer Customer Competition

Lower rates can make it cheaper for companies to run coding assistants and other AI services, particularly where usage is heavy. Reported customer-specific offers add another layer: buyers may be able to negotiate below published prices, while vendors compete to secure large contracts and ongoing usage.

But a lower per-token price does not automatically mean a lower total bill. Cached-token rates, included usage, contract duration and seat counts can all change the effective cost. ThorstenMeyerAI.com reports that Anthropic added capped, expiring API credits to its Max and Team plans, while its Enterprise plan requires at least 20 seats. It also reports that OpenAI reduced the usage included in new $200 Pro subscriptions without changing the subscription price.

That makes contract details and workload testing important for buyers. A company comparing two providers needs to measure what its own applications consume and whether a newer model performs adequately on its tasks. The reported price competition may improve bargaining power, but it does not establish that every customer can obtain the same discount or savings.

How Model Prices Shifted Since June

The price movement spans several model launches and revisions. Anthropic introduced Sonnet 5 in June at an introductory price of $2 per million input tokens and $10 per million output tokens, then made that pricing permanent in August, according to the source material. OpenAI cut Luna pricing in July and announced a temporary Sol price reduction in August before releasing GPT-6 models in September.

Pricing for cached tokens has also become part of the competition. These tokens can arise when an AI system reuses earlier prompt material. Data scientist Max Woolf estimates cached tokens account for 96% to 99% of token spending in agentic coding workloads. That estimate highlights why cache pricing may affect some coding customers more than standard input and output rates; it should not be read as a universal share for every workload.

Many of the lower prices apply to successor models rather than the same model becoming cheaper. OpenAI has described GPT-6.1 Sol as offering “near-Astra intelligence for a fifth of the price,” a company claim about capability and price. Customers would need to test their own tasks to determine whether a replacement model meets their requirements.

“Anthropic models handled 57% of the company’s coding work in May, while OpenAI handled 16%; the reported current shares are 18% for Anthropic and 52% for OpenAI.”

— Parker Conrad, Rippling chief executive, as reported by ThorstenMeyerAI.com

What Discounts and Savings Cover

The public reporting does not provide complete contract terms for the cited customer offers, including how long discounts last, what usage they cover or what conditions apply. It is also unclear whether the bank’s reported seat pricing and expanded allowance are available to other customers.

The 80% to 90% savings figure is attributed to DataCamp’s chief executive and concerns expected total costs relative to the company’s current setup. It is not a published OpenAI discount rate. The Information’s article is paywalled, and the source account describes relying on its public summary and other write-ups, so details beyond those accounts remain difficult to verify independently.

Nor do the pricing comparisons establish that newer models deliver equivalent results for every task. The actual impact on a customer’s bill depends on its usage mix, chosen models, caching, plan allowances and negotiated contract. The reported changes do not show whether price cuts will continue or how competitors will respond.

Compare Contracts and Workloads

Companies considering renewals or new AI contracts will need to compare their current usage with proposed terms, including per-token rates, cache charges, seat minimums, credits and expiration rules. Testing more than one model on the same representative tasks can help buyers judge whether lower-priced successors meet their quality and reliability needs.

Further pricing announcements, customer disclosures and contract renewals may clarify how broadly the discounts are available. For individuals on monthly plans, the near-term issue is whether usage allowances or model access change even when the subscription fee stays the same. Until providers publish fuller terms, buyers should treat reported savings as customer-specific and check the limits attached to their own plan.

Key Questions

Did OpenAI cut prices by 90%?

Not as a general list-price reduction. The 80% to 90% figure attributed to DataCamp’s CEO describes expected total savings against the company’s existing setup, according to reporting cited by ThorstenMeyerAI.com.

Which OpenAI prices fell?

The source reports GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, down from $5 and $30 for the earlier Sol model in June. Luna pricing also fell, but the comparison involves different model versions.

Has Anthropic also lowered prices?

Yes. The source reports Opus pricing fell from $5 and $25 per million input and output tokens for Opus 5 to $4 and $20 for Opus 5.5. Anthropic has also changed cache-read pricing for some models.

Why should businesses check more than list prices?

Total costs also depend on usage allowances, cached-token rates, seat minimums, credits and contract terms. A newer model’s performance on a company’s specific work can affect the value of a lower price.

Are the reported discounts available to every customer?

The available reporting describes particular customer offers and does not establish that the same terms are available to all buyers. Contract details and eligibility remain unclear.

Source: ThorstenMeyerAI.com