DeepSeek
Chinese AI Lab releasing open-weight frontier models at a fraction of US training cost.
DeepSeek, the Chinese AI Lab whose DeepSeek-R1 model reset assumptions about frontier AI training costs, has been outperformed on coding benchmarks for two consecutive years by the UK's Cosine, named a Sovereign AI Fund investee on 16 April 2026.
Last refreshed: 22 September 2026 · Appears in 2 active topics
How did DeepSeek change the calculus for European sovereign AI?
Timeline for DeepSeek
Recruited at the AI hiring track
AI: Jobs, Power & Money: Mentioned in: Beijing's AI job plan sets no targetsMentioned in: UK names first Sovereign AI investees
European Tech SovereigntyBackground
DeepSeek is a Chinese AI research laboratory backed by the quantitative hedge fund High-Flyer, based in Hangzhou. It released the open-weight DeepSeek-R1 and DeepSeek-V3 models in late 2024 and early 2025, benchmarking at or above GPT-4-class performance at a reported training cost far below comparable US frontier models. The models are released under permissive licences allowing commercial use and fine-tuning.
DeepSeek's release reset industry assumptions about the capital and compute needed to reach frontier AI performance. For European policymakers it carries a double signal: proof that frontier AI does not require US-scale data centre spending, and proof that a non-Western actor can compete at the frontier, a combination that has sharpened the case for Sovereign AI investment in the UK and EU.
The EU AI Act's general-purpose AI obligations apply to DeepSeek models deployed in the EU market from 2 August 2026, and on Independent coding benchmarks the UK's Cosine has outperformed DeepSeek for two consecutive years.
What is DeepSeek AI?
Why does DeepSeek matter for European AI sovereignty?
Is DeepSeek subject to the EU AI Act?
Can DeepSeek beat Cosine on coding benchmarks?
Written with AI from Lowdown's published reporting. Not individually reviewed by an editor. Updated 22 September 2026. Editorial standards.