DeepSeek

DeepSeek is a conversational AI assistant from the Chinese lab of the same name. You can talk to it for free at chat.deepseek.com, or build on top of it through a metered API that charges per million tokens. It handles general chat, coding, reasoning, and long-document work.

The current generation is DeepSeek-V4, whose preview launched on April 24, 2026 and replaced the older V3/R1 family. It ships in two tiers: V4-Pro (1.6 trillion total parameters, 49 billion active) aimed at competing with top closed-source models, and the cheaper default V4-Flash (284 billion total, 13 billion active). The legacy deepseek-chat and deepseek-reasoner model names are being deprecated on July 24, 2026.

Key capabilities include a new DeepSeek Sparse Attention (DSA) mechanism with token-wise compression for better efficiency, a standard 1M-token context window with output up to 384K tokens, and explicit agentic optimizations, including integration with agent tools such as Claude Code.

On pricing, the web chat is free. The API is paid, per 1M tokens: V4-Flash costs $0.14 input on a cache miss ($0.0028 on a cache hit) and $0.28 output; V4-Pro costs $0.435 input on a cache miss ($0.003625 on a cache hit) and $0.87 output. The legacy names map to V4-Flash's non-thinking and thinking modes.

Limitations to keep in mind: the API is a metered, paid service rather than free access, some third-party reports of a free token grant for new accounts are not confirmed in official documentation, and as a China-based service it may raise data-residency or compliance questions for some organizations.

DeepSeek suits developers who want a strong, low-cost model API, and casual users who want a capable free web assistant without a subscription.