DeepSeek API Prices Rise Up to 1,100% on Sunday, and Peak Hours Come With Them

DeepSeek will raise API prices for its V4 model family by up to 1,100% starting 16 August at 16:00 UTC, with new peak and off-peak pricing tiers. V4-Flash cache-miss input tokens rise from $0.14 to $0.44 per million at peak, and output tokens from $0.28 to $1.32; V4-Pro, which reached general availability this week, rises from $0.435/$0.87 to $1.32/$3.96 at peak. The company, which closed a first outside funding round topping $7 billion and is preparing for an IPO, said the split is meant to 'allocate resources more reasonably' as serving capacity, not demand, now sets prices.

The cheapest frontier API in the business just repriced itself around scarce compute. DeepSeek API prices go up on Sunday, and this is nothing like the modest tweak that phrase usually implies. From 16:00 UTC on 16 August, the company’s V4 family moves to a rate card that lands somewhere between roughly 50% and more than 1,100% above what developers pay today, depending on the model, the token type and the hour of the day. The headline figures are easy enough to follow. V4-Flash currently bills $0.14 per million cache-miss input tokens and $0.28 per million output tokens, flat, around the clock. On Sunday that becomes $0.44 and $1.32 at peak, halved off-peak. V4-Pro, the flagship that only reached general availability this week, moves from $0.435 / $0.87 to $1.32 / $3.96 at peak, or $0.66 / $1.98 outside those windows. The four-figure percentage in the headlines comes from cache reads, the discount that made DeepSeek almost free for agents replaying the same long context over and over. An API with peak hours The structural change matters more than any single number. DeepSeek is carving the day into peak and off-peak windows: peak covers 01:00 to 04:00 and 06:00 to 10:00 UTC, and everything else bills at half rate. The company said the split is meant to “allocate resources more reasonably” https://www.engadget.com/2236912/deepseek-ai-models-get-four-times-pricier/ , which is a polite way of saying serving capacity, not demand, is the thing setting the price now. That framing invites an easy mistake, so it’s worth being blunt about it. Off-peak is not a discount against today’s bill. It is half of a raised peak, and even the cheapest new tier sits above the old flat rate. Nobody’s costs go down on Sunday. Cheap tokens were a hardware story all along DeepSeek’s whole reputation was built on undercutting Western labs by something close to an order of magnitude, which is exactly why a Chinese lab with no access to top-end Nvidia silicon rattled the market in the first place. That gap doesn’t vanish here — even after Sunday, V4-Pro remains well under what the big US frontier APIs charge. But the direction of travel is new, and it is the first time one of these providers has openly told customers that when they run a job is a pricing input. There’s a commercial layer to it too. Bloomberg https://www.bloomberg.com/news/articles/2026-08-13/deepseek-increases-prices-for-ai-services-by-multiple-times reported the increase brings DeepSeek’s rates closer to its rivals’, and the company is laying groundwork for an IPO after closing a first outside funding round that topped $7 billion. Both things can be true at once: a company heading for public-market scrutiny needs margins, and a company renting scarce accelerators cannot keep losing money on every cached token. Hardware people will recognise the pattern. The same squeeze that has DDR5 kits selling for four times their 2025 price, and that pushed both current consoles into mid-generation price hikes, is now showing up in the cost of an API call. Memory, packaging and accelerator capacity are all spoken for, and every layer above them is quietly repricing to match. If you build on DeepSeek, the practical response is dull but effective: find out how much of your monthly spend is cache reads before Sunday, because that line is about to move the most; push batch and evaluation runs into off-peak hours; and price any customer-facing product you’ve been running on $0.28 output tokens as though it costs five times that, since shortly it will. More detail on the new rates is in InfoWorld’s breakdown https://www.infoworld.com/article/4209439/deepseek-raises-some-v4-prices-by-more-than-10x-as-ai-demand-strains-capacity.html and Caixin’s report https://www.caixinglobal.com/2026-08-14/deepseek-launches-v4-pro-and-raises-api-prices-by-as-much-as-1100-102473919.html .