DeepSeek AI review: the free chatbot that shocked Silicon Valley

What is DeepSeek?

DeepSeek is a Chinese AI company whose chatbot and models shocked the industry in early 2025 by matching the performance of the leading Western models — while reportedly costing a fraction of the money to train. It's a general-purpose chat assistant: ask questions, draft text, write code, reason through problems.

A short history

• Founded in 2023 by Liang Wenfeng, spun out of the quantitative hedge fund High-Flyer.
• Released DeepSeek-V2 in 2024, quickly gaining a reputation for efficiency.
• DeepSeek-R1 (early 2025) was the shock: a reasoning model that rivalled top-tier systems at a claimed fraction of the training cost — it briefly wiped billions off US tech stocks.
• Its open-weight models (DeepSeek-V3, R1) are freely downloadable, which made it hugely popular with developers.

What it's great at

• Reasoning and maths — its R-series models 'think step by step' and are strong on logic problems.
• Coding — one of the best free options for writing and debugging code.
• Cost — genuinely free to use, and cheap for developers via API.
• Open weights — developers can run it themselves.

What to watch out for

• Privacy — it's a Chinese company; avoid feeding it sensitive or personal data.
• Some governments and workplaces have restricted it on security grounds.
• Availability can be patchy at peak times.

Verdict

If you want a powerful free chatbot for general tasks, reasoning, or coding, DeepSeek is genuinely excellent. Just keep sensitive information out of it. For everyday 'explain this to me' and 'draft this for me' work, it holds its own against anything.