DeepSeek
Hangzhou-based open-weights frontier lab; spun from quant fund High-Flyer, now reportedly raising its first external round
PRIVATEHangzhou, China~150 (estimated)Founded 2023Historically self-funded by High-Flyer; first external round reported 2026 raised$52-59B (reported, first external round) valuationwww.deepseek.com
SectorsAI Software
Market Metrics
Funding (reported 2026)
~$7.4B / $52-59B
first external round, led by Tencent + CATL; previously self-funded
V4-Pro (Apr 2026)
1.6T / 49B active
near GPT-5.5 / Opus 4.7 at far lower API cost; 1M context
US App Store Rank (peak)
#1 free
Jan 28 2025; first Chinese consumer AI app to hit #1 in the US
NVDA Market-Cap Impact (Jan 27 2025)
-$589B / -17%
Largest single-day market-cap loss for any US-listed company in history
Updated at 2026-Q2
Funding History
Total raised: Historically self-funded by High-Flyer; first external round reported 2026 · Latest valuation: $52-59B (reported, first external round)
| Round | Date | Amount | Valuation | Key Investors |
|---|---|---|---|---|
| Founder / Parent Funding | 2023-2025 | Undisclosed | - | High-Flyer Quantitative Investment Management |
| First External Round (reported) | 2026 (reported) | ~$7.4B (50B yuan) target | $52-59B (reported) | TencentCATLLiang Wenfeng (~40% reported)state-linked chip fund (reported) |
Updated at 2026-Q2
Product Specifications
| Product | Type | Status |
|---|---|---|
| DeepSeek-V4 | Open-weights MoE (V4-Pro 1.6T/49B, V4-Flash 284B/13B, 1M context) | Released (Apr 2026) |
| DeepSeek-V3 | 671B parameter MoE open-weights base model (37B active) | Released (Dec 2024) |
| DeepSeek-R1 | Open-weights reasoning model (chain-of-thought) | Released (Jan 2025) |
| DeepSeek-Coder family | Open-weights code models (1.3B to 33B) | Released |
| DeepSeek Chat | Consumer chatbot on web + iOS + Android | GA |
| DeepSeek API | Developer API at heavily discounted price-per-token | GA |
Updated at 2026-Q2
Partnerships & Investors
High-Flyer QuantParent / Founding Funder
Hedge fund founded 2015 by Liang Wenfeng; banked thousands of A100s and H800s pre-export-controls; funded DeepSeek's first years entirely from its balance sheet.
Tencent + CATLReported Lead Investors (2026)
Reported lead cheques in DeepSeek's first external round at a $52-59B valuation.
Tencent + Baidu + Huawei + ByteDanceDistribution (PRC)
DeepSeek models hosted across all major Chinese clouds; Huawei Ascend 910C / inference pipeline.
Beijing Academy of AI (BAAI)Research Adjacency
DeepSeek researchers frequently co-publish with BAAI; loose alignment with PRC national AI strategy.
Updated at 2026-Q2
Key Milestones
Jul 2023
Company Founded
Liang Wenfeng spins DeepSeek out of his quant hedge fund High-Flyer in Hangzhou; uses banked A100 / H800 capacity for training
Nov 2023
DeepSeek LLM (v1)
First open-weights release (7B and 67B parameter base models)
May 2024
DeepSeek-V2
236B parameter MoE; Multi-head Latent Attention (MLA) paper published
Dec 2024
DeepSeek-V3
671B parameter MoE (37B active); claimed total training cost ~$5.6M / 2.78M H800 GPU-hours (figure heavily disputed)
Jan 20 2025
DeepSeek-R1
Open-weights reasoning model competitive with OpenAI o1; chain-of-thought training pipeline published
Jan 27 2025
Market Shockwave
DeepSeek app hits #1 free on US App Store; NVIDIA falls 17% / -$589B market cap in one session, the largest single-day market-cap loss in history
Feb 2025
Xi Jinping Symposium
Liang Wenfeng among small group of CEOs meeting Xi Jinping; signals Beijing's elevation of DeepSeek as a national champion
Mar 2025
Western Government Bans
Italy, Australia, Taiwan, US House, NY State, several US federal agencies restrict DeepSeek on government devices
Apr 24 2026
DeepSeek-V4
V4-Pro (1.6T total / 49B active) and V4-Flash (284B / 13B), both with 1M-token context; benchmarks near GPT-5.5 / Claude Opus 4.7 at far lower API cost; optimized for Huawei Ascend chips
Jun 2026
First External Round (reported)
Reportedly raising ~$7.4B (50B yuan) at a $52-59B valuation led by Tencent and CATL, the lab's first outside capital
Updated at 2026-Q2