DeepSeek
Open-weight model lab with a chat app and very low-cost API
DeepSeek researches general-purpose large models and has open-sourced models such as DeepSeek-V4 and DeepSeek-R1. It offers a chat app plus a developer API billed per token, with a 1M-token context window and both OpenAI- and Anthropic-compatible endpoints, so many coding tools can plug in directly. A strong choice if you want capable reasoning at low cost.
What it's good for
- Writing, analysis and reasoning questions
- Power coding tools as the backend model
- Batch-process text through the API
- Self-host the open-weight models
Pricing
API is billed per 1M tokens, e.g. deepseek-flash input (cache miss) $0.30 peak / $0.15 off-peak and output $1.20 peak / $0.60 off-peak; deepseek-v4-pro costs more. Off-peak is half the peak rate. Check the site for chat app pricing.
Access from mainland China
The site carries a Chinese ICP filing and is directly reachable in mainland China.