DeepSeek cuts V4‑Pro price 75%

DeepSeek said on May 24 that it would make a 75% discount on its V4-Pro model permanent, extending pricing that had been scheduled to expire on May 31. The move keeps the Chinese startup’s flagship model at one-quarter of its original launch price and adds pressure in a year of falling AI API costs. DeepSeek’s official pricing documentation says the change will leave V4-Pro at $0.435 per million input tokens and $0.87 per million output tokens after the promotion ends. ### What exactly changed in DeepSeek’s pricing? DeepSeek’s API pricing page says V4-Pro pricing will be “officially adjusted to 1/4 of the original price” after the 75% promotion ends on May 31 at 15:59 UTC. The same page says the original V4-Pro rates were $1.74 per million input tokens and $3.48 per million output tokens, implying the permanent post-promotion rates are $0.435 and $0.87. (api-docs.deepseek.com) The company also says cache-hit input pricing across models was cut to one-tenth of launch pricing effective April 26, a separate reduction that lowers costs further for applications that reuse prompt context. DeepSeek’s pricing materials list V4-Pro cache-hit input pricing at about $0.0036 per million tokens. ### Why are developers paying attention to this one model? (api-docs.deepseek.com) DeepSeek launched V4-Pro and V4-Flash on April 24 and described V4-Pro as its higher-end model for reasoning, coding and agent-style tasks. The company’s release said V4-Pro supports a 1 million-token context window, is available through the same API base URL as earlier models, and works with both OpenAI ChatCompletions and Anthropic-style interfaces. (api-docs.deepseek.com) OpenRouter’s model page lists the same current V4-Pro API rates — $0.435 per million input tokens and $0.87 per million output tokens — and says the model has a 1,048,576-token context window with a maximum output of 384,000 tokens. ### Is this only a pricing story, or also a compatibility story? DeepSeek’s documentation says developers do not need to change the base URL to use V4-Pro or V4-Flash and only need to update the model parameter. (api-docs.deepseek.com) The company also says the API supports both OpenAI and Anthropic interfaces, which lowers migration work for teams already using those formats. (openrouter.ai) Techmeme’s roundup of the announcement described the price cut as part of a broader AI cost fight, summarizing DeepSeek’s move as a reduction to $0.435 per million input tokens and $0.87 per million output tokens. The Next Web separately reported that the permanent discount escalates competition in model APIs. ### How does this compare with where V4-Pro started? (api-docs.deepseek.com) DeepSeek’s April 24 release introduced V4-Pro as part of a new V4 family and positioned it against top closed-source models on reasoning and coding tasks. Third-party pricing guides published around launch listed V4-Pro at $1.74 per million input tokens and $3.48 per million output tokens before the discount, matching the quarter-price math in DeepSeek’s current pricing page. (techmeme.com) Times of India reported on May 25 that the company had decided to keep the 75% cut in place permanently. That report said the lower price would remain after the original May 31 deadline. (api-docs.deepseek.com) ### What comes next for users already on DeepSeek? DeepSeek’s API docs say the legacy model names “deepseek-chat” and “deepseek-reasoner” will be discontinued on July 24, 2026, and currently route to V4-Flash modes for compatibility. The company says developers who want the new flagship pricing and capabilities should call `deepseek-v4-pro` directly at the existing API endpoint. (api-docs.deepseek.com) (timesofindia.indiatimes.com)