DeepSeek V4 Pro pricing compared

DeepSeek’s published API pricing put its V4-Pro model well below two U.S. rivals on May 24, according to the company’s pricing page and official pricing pages from OpenAI and Anthropic. DeepSeek listed V4-Pro at $0.435 per million input tokens and $0.87 per million output tokens. OpenAI listed GPT-5.5 at $5 per million input tokens and $30 per million output tokens, while Anthropic listed Claude Opus 4.7 at $5 and $25. The comparison circulated on X on May 24, but the underlying prices are also shown on the companies’ own websites. ### What exactly did DeepSeek publish? DeepSeek’s API documentation lists “deepseek-v4-pro” with pricing in units of per 1 million tokens. The page shows $0.435 for cache-miss input tokens, $0.003625 for cached input tokens and $0.87 for output tokens. It also lists a 1 million-token context length and a 384,000-token maximum output. (api-docs.deepseek.com) DeepSeek’s pricing page also says the V4-Pro API price “will be officially adjusted to 1/4 of the original price after the 75% discount promotion ends on 2026/05/31 15:59 UTC.” That means the promotional level cited in social posts is, according to DeepSeek’s own documentation, set to become the official post-promotion rate rather than reverting to the older $1.74 input and $3.48 output levels shown on the same page. (api-docs.deepseek.com) ### How does that compare with OpenAI and Anthropic? OpenAI’s API pricing page lists GPT-5.5 at $5.00 per 1 million input tokens, $0.50 for cached input tokens and $30.00 per 1 million output tokens. OpenAI’s model page for GPT-5.5 repeats those token prices and notes that prompts above 272,000 input tokens are billed at higher rates for the full session. (api-docs.deepseek.com) Anthropic’s Claude Opus 4.7 page says pricing starts at $5 per million input tokens and $25 per million output tokens. Anthropic’s April 16 product announcement said Opus 4.7 pricing remained the same as Opus 4.6 at those levels. Using those posted rates, DeepSeek’s listed V4-Pro input price is about 91% lower than GPT-5.5’s and Claude Opus 4.7’s, and its output price is about 97% lower than GPT-5.5’s and about 96.5% lower than Opus 4.7’s. (openai.com) Those percentages are arithmetic comparisons based on the companies’ published per-million-token prices. ### Why did this show up on social media? (anthropic.com) An X post shared on May 24 highlighted the gap between DeepSeek’s published V4-Pro prices and the posted prices for GPT-5.5 and Claude Opus 4.7. X did not return readable page text in this search session, so the social post itself could not be quoted directly here, but the price figures referenced in the post match the companies’ official pricing pages. (api-docs.deepseek.com) ### Are these prices directly comparable? The three companies all present token pricing on official product or API pages, but the pages also show differences in billing details. DeepSeek distinguishes cache-hit and cache-miss input prices and warns that product prices may vary. OpenAI lists separate standard and batch-related pricing details and adds higher-session pricing for very large GPT-5.5 prompts. (api-docs.deepseek.com) Anthropic says Opus 4.7 pricing can be reduced with prompt caching and batch processing and adds a 1.1x surcharge for U.S.-only inference. ### What is the next dated milestone to watch? May 31, 2026 at 15:59 UTC is the next specific date on DeepSeek’s pricing page. DeepSeek says that is when the 75% discount promotion ends and when V4-Pro pricing is to be officially adjusted to one-quarter of the original rate. (api-docs.deepseek.com)