DeepSeek

Overview

To pursue artificial general intelligence through open, research-driven development.

Founded in 2023 by Liang Wenfeng, who also co-founded the quantitative hedge fund High-Flyer, from which DeepSeek was spun out as an independent lab.

Flagship AI models

Key models released by DeepSeek, most recent last. Dates are the public release month where confirmed.

1. 2024-12
   DeepSeek-V3671B-parameter mixture-of-experts base model
2. 2025-01
   DeepSeek-R1Open reasoning model that matched OpenAI o1 on several benchmarks
3. 2026-04
   DeepSeek-V4Latest generation with native long-context support

Funding & financials

Quantitative figures from Epoch AI's Data on AI Companies (CC BY 4.0). Each is the latest dated figure Epoch holds; private labs disclose selectively, so undisclosed values are marked as such — never estimated.

Key facts

• The January 2025 release of DeepSeek-R1 was widely described as an 'AI Sputnik moment' for demonstrating frontier reasoning at a fraction of typical training cost.
• DeepSeek's technical report estimated roughly USD 5.6 million of compute to train the underlying V3 model.
• Ships models under the permissive MIT licence, allowing downloading, fine-tuning and commercial use of the weights.
• Backed by the resources of quantitative hedge fund High-Flyer, which supplied its GPU compute.

Frequently asked

When was DeepSeek founded?

DeepSeek was founded in 2023, by Liang Wenfeng. It is headquartered in Hangzhou, China.

How much funding has DeepSeek raised?

DeepSeek has raised about $7.5bn in total equity funding, at a latest valuation of about $50.0bn (2026-06-16). Source: Epoch AI.

What are DeepSeek's main AI models?

DeepSeek's notable models include DeepSeek-V3, DeepSeek-R1, DeepSeek-V4. Cost-efficient open-weight reasoning models; China's leading open-model lab.

Sources

• DeepSeek
• Wikipedia — DeepSeek
• Epoch AI — Notable AI Models
• Epoch AI — Data on AI Companies (financials, CC BY 4.0)

Other AI labs

← All AI labs