Anthropic prices Opus 4.7 at $5/$25

Anthropic’s latest pricing update adds a new marker in the AI model cost race. On May 24, Anthropic’s Claude model documentation listed Opus 4.7 at $5 per million input tokens and $25 per million output tokens, according to reporting by The Next Web and Anthropic’s model overview page. DeepSeek, by contrast, has already published lower prices for its V4-Pro model and says another cut is due after its current promotion ends on May 31. The result is a wider spread between premium U.S. frontier-model pricing and the rates some Chinese rivals are now advertising. ### Where did the $5/$25 figure come from? The Next Web reported on May 24 that Anthropic had updated its Claude documentation to show Opus 4.7 priced at $5 per million input tokens and $25 per million output tokens. Anthropic’s model overview documentation also reflected an updated Claude lineup on Sunday, though the search results available did not expose the full pricing table directly. The published numbers matter because API buyers are usually billed separately for tokens sent into a model and tokens generated back. At Anthropic’s listed rates, output remains five times more expensive than input, a structure common in premium reasoning and long-form generation models. ### How far above DeepSeek is that? DeepSeek’s API pricing page says V4-Pro will be “officially adjusted to 1/4 of the original price” after its 75% discount promotion ends on 2026/05/31 15:59 UTC. (thenextweb.com) The same documentation says the adjustment follows a promotion that began on April 26. The Next Web reported earlier that even before the promotional discount, DeepSeek-V4-Pro cost $0.145 per million input tokens and $3.48 per million output tokens. (thenextweb.com) Those levels were already below Anthropic’s Opus 4.7 pricing, according to the publication’s comparison. ### What exactly did DeepSeek change this month? DeepSeek announced on April 26 that it was offering a 75% discount on V4-Pro and cutting input cache-hit prices across its API suite to one-tenth of previous levels, according to The Next Web’s April 27 report. (api-docs.deepseek.com) The company’s API documentation now says the V4-Pro model’s standard pricing will still settle at one-quarter of the original level once the promotion ends. (thenextweb.com) DeepSeek’s documentation also shows V4-Pro and V4-Flash as current models in its API and says older model names, deepseek-chat and deepseek-reasoner, will be discontinued on July 24, 2026. ### Why are developers watching token prices so closely? API pricing has become a direct operating-cost question for companies building coding tools, agents and enterprise workflows on top of foundation models. (thenextweb.com) Every prompt, retrieval step and generated answer adds token usage, which means differences of a few dollars per million tokens can compound quickly at scale. That is one reason price tables from Anthropic, DeepSeek and other model providers are now watched as closely as benchmark charts. (api-docs.deepseek.com) The comparison is not only about raw price. Model buyers also weigh latency, context length, reliability and tool use, and providers often package those trade-offs differently. But the published numbers give developers a clear first screen when deciding which models to test and which workloads to route elsewhere. ### What happens next? May 31 is the next concrete date in this pricing story. DeepSeek’s API docs say its 75% V4-Pro promotion ends on May 31, 2026, at 15:59 UTC, when the model’s standard price is due to reset to one-quarter of its original level. Anthropic’s current Claude lineup remains posted in its model overview documentation, which is where developers will look for any further pricing changes. (api-docs.deepseek.com)