{"rewrite":{"id":"r_1bc0c15301c7415fb5b3e95c","clusterId":"c_336f13aa6e194f3e98d11038","slug":"deepseek-makes-75-api-discount-permanent-on-v4-pro-model","model":"deepseek-v4-flash","headline":"DeepSeek Makes 75% API Discount Permanent on V4-Pro Model","summary":"Chinese AI startup DeepSeek announced on May 22 that it will permanently apply a 75% discount to API pricing for its V4-Pro model, extending a limited-time promotion from late April. New rates are $0.435 per million input tokens and $0.87 per million output tokens, undercutting OpenAI and Anthropic models. Security concerns persist over Chinese server data processing.","whyItMatters":"The permanent price cut positions DeepSeek as a cost-disruptive force in the AI API market, challenging established US providers while the security debate around Chinese-hosted models continues.","webCardHtml":"\u003cp\u003eDeepSeek made its V4-Pro API discount permanent on May 22, dropping prices to $0.435 per million input tokens and $0.87 per million output tokens. The move extends a promotion that began in late April and puts output pricing well below OpenAI's GPT-5.5 and Anthropic's Claude Opus series. The V4 series is known for 1 million token long-context support and strong math and coding performance, but US security concerns about data processed on Chinese servers remain. Developer reactions on Reddit included surprise at the sub-$1-per-million-token rate and questions about profitability.\u003c/p\u003e","blueskyPost":"DeepSeek's permanent 75% discount on V4-Pro brings output pricing to $0.87 per million tokens, undercutting GPT-5.5 and Claude Opus. The margin question is whether volume can offset the rate.","twitterPost":"DeepSeek's V4-Pro permanent discount undercuts GPT-5.5 and Claude Opus on output pricing. The unanswered question is unit economics at these rates.","threadsPost":"DeepSeek made its 75% V4-Pro API discount permanent, dropping output to $0.87 per million tokens. That is cheaper than GPT-5.5 and Claude Opus. Developer reaction is split between enthusiasm and skepticism about profitability. The security concerns around Chinese server data processing remain unresolved.","newsletterBlurb":"DeepSeek permanently applied a 75% discount to its V4-Pro API pricing, bringing rates to $0.435 per million input tokens and $0.87 per million output tokens. The move undercuts OpenAI and Anthropic models but keeps security concerns about Chinese server data processing in focus.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/404/4404606/?rss\",\"title\":\"\\\"Too Cheap\\\": DeepSeek's New Prices Surprise, Security Concerns Also Draw Attention\"}]","lintFlagsJson":"[]","lintHits":0,"costUsd":0,"inputTokens":4208,"outputTokens":548,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1779826512,"createdAt":"2026-05-25T02:02:19.000Z","publishedAt":"2026-05-25T02:16:03.000Z","updatedAt":"2026-05-25T02:16:03.000Z"},"cluster":{"id":"c_336f13aa6e194f3e98d11038","canonicalTitle":"「安すぎる」DeepSeek新価格に驚きの声　セキュリティ懸念も注目集める","representativeArticleId":"a_b74483b2694d23bbdd261225","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":false,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-05-25T00:00:00.000Z","lastSeenAt":"2026-05-25T09:30:00.000Z","updatedAt":"2026-05-25T11:23:31.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/404/4404876/?rss","title":"OrcaRouter、DeepSeek V4 Pro APIを75%割引価格で提供開始 ― エンタープライズAIエージェントワークフローのコスト最適化を加速"}],"entities":{"anime_titles":[],"manga_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":["On May 22, DeepSeek permanently applied a 75% discount to its V4-Pro API pricing, reducing rates to $0.435 per million input tokens and $0.87 per million output tokens.","The discount extends a promotion that began in late April, making output pricing lower than OpenAI's GPT-5.5 and Anthropic's Claude Opus series.","The V4-Pro model supports a 1 million token context window and is known for strong math and coding performance.","Security concerns persist over data processed on Chinese servers, as noted in the article.","Developer reactions on Reddit included surprise at the sub-$1-per-million-token rate and questions about profitability."]}