China AI DataTokenHeart News ↗
Foundation Models

DeepSeek

深度求索

Open-weight frontier models built at a fraction of the usual cost

Headquarters
Hangzhou, China
Region
Hangzhou
Founded
2023
Stage
Undisclosed
01

Overview

DeepSeek is a Hangzhou-based AI lab incubated by the quantitative hedge fund High-Flyer. Its open-weight models rival top proprietary systems while training at far lower cost. The V4 generation, released in 2026 under an MIT license, is among the largest open-weight models available and is text-only.

02

Products

2

DeepSeek-V4-Pro

Flagship open-weight MoE model (~1.6T params), 1M-token context

DeepSeek-V4-Flash

Faster, lighter V4 variant for high-volume use

03

Highlights

2
  • 01V4 (2026) is among the largest open-weight models, released under a permissive MIT license
  • 02Frontier-level performance at a fraction of typical training cost; widely used by developers worldwide
04

People

1
06

Industry chain

1

Upstream suppliers

DeepSeek's operator, High-Flyer, purchased thousands of NVIDIA chips for training DeepSeek's early models. The cited text contains an apparently erroneous year, so the purchase date cannot be established from this excerpt.

Source

Downstream customers

Not yet listed

Competitors

Not yet listed

07

Funding

0

Not yet listed

08

Events

0

Not yet listed

09

Related projects

0

Not yet listed

10

Topics

5
11

Coverage

11

Information compiled from public sources; refer to the company's official website for authoritative details.