DeepSeek Secures $10B Funding and Slashes API Prices by 75%

DeepSeek announced that its API price will be reduced to one‑quarter of the original price, a permanent 75% discount, making its rate lower than GPT‑5.5 and Claude Opus 4.7 and switching the unit from USD to CNY.

DeepSeek V4 Pro maintains world‑leading model performance while its novel attention architecture, parallel strategy and memory quantization cut inference cost to less than one‑tenth of comparable international models, which the author says makes it an almost forced choice for developers needing high‑frequency API calls.

The price cut sparked widespread excitement and raised doubts about sustainability; blogger Yuchen Jin suggested the move could stem from a breakthrough in inference optimization or lower Huawei chip costs.

Bloomberg reported that DeepSeek is raising up to 700 billion CNY (~$10 billion), which would set a record for a Chinese tech startup’s first‑round financing; investors include the National AI Industry Investment Fund, Tencent, IDG Capital, Monolith Capital, and possibly CATL, JD, NetEase, with founder Liang Wenfeng also possibly contributing 200 billion CNY.

Founder Liang Wenfeng emphasized that advancing AI research and ultimately achieving AGI take precedence over short‑term monetization, and that the company will continue developing open‑source models.

DeepSeek has launched a Harness team to productize its technology, directly targeting Anthropic’s Claude Code; the team will build a desktop‑agent product where Model + Harness = Agent, aiming to turn V4’s code‑generation capability into a seamless developer workflow.

The simultaneous price cut, massive funding, and productization push suggest DeepSeek is positioning itself to dominate the “AI programmer” scenario and marks a turning point for the Chinese AI industry.