DeepSeek makes the V4 Pro price discount permanent

DeepSeek’s huge price cut is here to stay, and users are acting like they found cheat codes

TLDR: DeepSeek made its big V4 Pro discount permanent, turning a temporary sale into a lasting price cut that could shake up the market. Commenters are split between cheering the absurdly cheap costs, bragging about strong results, and asking suspicious questions about hardware and data privacy.

DeepSeek just turned a temporary sale into the new normal, announcing that its V4 Pro pricing will stay at one-quarter of the original cost after the promo ends. It also slashed some other charges, including cached input prices, and quietly warned that older model names will eventually be phased out. But let’s be real: the paperwork isn’t the juicy part. The real fireworks are in the community, where people are reacting like they just stumbled onto the discount aisle of the future.

The loudest vibe by far is shock at the value. One user called it “insane value” and casually dropped the kind of brag that makes rivals sweat: after using 65 million tokens, they say it only cost them $1.5. That sparked the biggest mood of the thread: DeepSeek isn’t just cheap, some users think it’s actually better than competing tools for difficult coding work. Another commenter praised it as the first model with a usable giant memory window, basically saying, finally, something affordable that doesn’t tap out halfway through the job.

But because this is the internet, the celebration came with side-eye. One person popped in with the spicy little grenade, “is this the Huawei chip difference?”—the kind of comment that instantly turns a pricing thread into a geopolitics-adjacent conspiracy lounge. Meanwhile, another user skipped the hype entirely and went straight into workplace paranoia mode, asking if there’s a safe no-retention middleman for companies pushing hundreds of millions of tokens a day. So yes, people love the bargain—but they also want to know what the catch is, who’s powering it, and whether their boss will approve it. Classic comment-section chaos.

Key Points

  • DeepSeek prices its API usage per 1 million tokens and bills based on the total number of input and output tokens.
  • The model names deepseek-chat and deepseek-reasoner are slated for future deprecation and currently map to non-thinking and thinking modes of deepseek-v4-flash.
  • For all models, the input cache hit price was reduced to one-tenth of the launch price effective 2026/4/26 12:15 UTC.
  • DeepSeek said deepseek-v4-pro pricing will be permanently adjusted to one-quarter of its original price after the 75% discount promotion ends on 2026/05/31 15:59 UTC.
  • Charges are deducted from topped-up or granted balances, with granted balance used first, and DeepSeek notes that prices may change over time.

Hottest takes

“insane value” — Sphax
“First accessible model with useable 1 million context window” — bel8
“is this the Huawei chip difference?” — kingjimmy
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.