DeepSeek explained: frontier reasoning, rock-bottom price.
DeepSeek is the open-weight lab that made everyone else cut prices. It delivers near-frontier reasoning and coding at roughly a tenth of the cost of the big US models — through a free chat app and a drop-in API. Here is what it is, why it matters, and how to start.
01What DeepSeek actually is
DeepSeek is a Chinese AI lab whose open-weight models combine strong reasoning with prices far below the frontier. You can use it three ways: the free chat app and website, a paid API that mirrors OpenAI and Anthropic's formats, or by running the open weights yourself. The headline is value — capability that lands close to the top models, at a cost that changes what is affordable.
02The lineup: V4 Flash and V4 Pro
Two models cover most needs. V4 Flash is fast and extremely cheap — the everyday workhorse. V4 Pro trades speed for deeper reasoning on hard problems. Both share a very large ~1M-token context window, so they can read entire documents or codebases in one go.
DeepSeek showed you could train and serve a near-frontier model for a fraction of the assumed cost — then published the weights. That combination of price and openness forced competitors to rethink pricing and put serious AI within reach of solo builders and small teams.
03Where it fits — and where it doesn't
Reach for DeepSeek when cost or openness matters: high-volume tasks, self-hosting, reasoning and coding on a budget. It is less of a fit when you need the absolute strongest model on a frontier task, the deepest multimodal features, or a vendor with Western enterprise compliance guarantees. For most everyday work, the gap is small and the savings are not.