August 13, 2026
New model, who dis-bill?
DeepSeek API Pricing Update
DeepSeek drops a shiny new model, and users instantly panic over the bigger bill
TLDR: DeepSeek launched its new V4-Pro system and changed pricing, including cheaper quiet-hour rates and higher busy-hour charges. The community immediately fixated on the cost jump, with some calling it a bad deal and others arguing DeepSeek had simply been too cheap before.
DeepSeek tried to serve good news and bad news in the same post: a brand-new V4-Pro model, upgraded tools, and a new pricing system with cheaper off-peak hours and pricier peak-time rates. But in the comments, nobody was popping champagne over the launch. The real mood was more like: "wait, how much now?"
The loudest reaction was sticker shock. One user bluntly called it "about 3x increase" and declared rival Luna the better bargain now, instantly turning the launch thread into a price-war cage match. Another commenter went even deeper into the weeds and warned that for people who rely on long-running coding assistants, one part of the bill could jump 6x off-peak and 12x at peak. Translation for normal humans: if you use this stuff heavily, your costs might suddenly feel a lot less cute.
That said, not everyone was grabbing a pitchfork. One of the more measured voices argued DeepSeek had been undercharging before, and even after the increase, it may still beat many competitors on some costs. So yes, there was outrage—but also a mini-countermovement of people saying, basically, "calm down, it was too cheap to last."
There was also some classic internet side-drama: users rushed in with direct pricing links and even mirror links to dodge the platform where the announcement was posted. Nothing says community trust issues like bringing receipts, archives, and alternate links to the party.
Key Points
- •DeepSeek announced the launch of DeepSeek-V4-Pro.
- •The V4 release includes major agent upgrades intended to improve production performance.
- •DeepSeek-V4-Pro and V4-Flash now support flexible reasoning effort levels for different task complexity.
- •The lineup adds native support for the OpenAI Responses API.
- •DeepSeek is introducing peak and off-peak API pricing, with off-peak rates 50% lower than peak starting at 16:00 UTC on Aug. 16, 2026.