DeepSeek is a Chinese AI lab that trains and open-weights frontier language models, known for reaching competitive benchmarks at unusually low training cost.
| date | round | amount | valuation | investors |
|---|---|---|---|---|
| May 2026 | Strategic | not disclosed | not disclosed | CATL |
DeepSeek's 500GB encoder-decoder model (8B+16B params) beats frontier models on design benchmarks while 20x faster and cheaper; its open-weight release forces Western labs to compete on cost and visual reasoning.