DeepSeek V4 Pro pricing just became a lot harder for rival AI labs to ignore.
DeepSeek now lists V4 Pro API prices at one quarter of the original launch price. The move effectively makes a 75 percent discount permanent after a promotion that had been scheduled to end on May 31.
On the pricing page, V4 Pro cache-hit input now costs $0.003625 per million tokens. Cache-miss input now costs $0.435. Output now costs $0.87. The old listed prices were $0.0145, $1.74, and $3.48 for the same token categories.
Why developers care
For casual users, those numbers may look tiny. However, AI agent workflows can burn through millions of tokens quickly. A lower output price can matter when a coding assistant, research bot, or document tool loops through long context windows all day.
DeepSeek also lists a 1 million-token context length and a 384K maximum output limit for V4 Pro. So the company is pairing the price cut with a pitch aimed at high-volume work, not just chat demos.
That matters for startups and solo builders. If model quality stays competitive, cheaper inference can let small teams test agents, retrieval apps, and long-document workflows without burning through a budget in a weekend.
It also gives procurement teams a stronger bargaining chip. Even companies that never deploy DeepSeek may use the new price floor during talks with other AI vendors.
The cut lands while every major AI company is trying to win developers. OpenAI, Google, Anthropic, and newer labs all need models that feel fast, capable, and affordable. We have seen the same pressure in app workflows like ChatGPT inside PowerPoint, where token costs can shape product decisions.
The caveat is trust. DeepSeek can undercut rivals on price, but enterprises still weigh privacy, reliability, compliance, and geopolitical risk. Even so, cheaper flagship inference gives the AI market one more reason to keep cutting prices.
