The short answer
DeepSeek API is billed per token. It has no free tier (5M free tokens on signup, valid 30 days, no card). Paid usage is quoted by the vendor rather than listed as a public price.
Plans and prices
| Plan | Price | What it covers |
|---|---|---|
| deepseek-v4-flash (input, cache miss) | $0.14 / 1M tokens | Cost-optimized model; cache-hit input just $0.0028/1M. Legacy deepseek-chat/reasoner alias to this model. |
| deepseek-v4-flash (output) | $0.28 / 1M tokens | Output tokens for V4 Flash, thinking and non-thinking modes. |
| deepseek-v4-pro (input, cache miss) | $0.435 / 1M tokens | Flagship reasoning model; cache-hit input $0.003625/1M. |
| deepseek-v4-pro (output) | $0.87 / 1M tokens | Output tokens for V4 Pro. 1M-token context, up to 384K output. |
How DeepSeek API bills you
The pricing model is per token. The price-leader for strong reasoning models, with OpenAI- and Anthropic-compatible endpoints and aggressive cache pricing. Best suited to: lowest-cost reasoning models.
What to watch for on the bill
- Dynamic, load-based rate limiting with no paid tier to raise limits; 429 errors spike under platform load
The pricing page we read
Source: https://api-docs.deepseek.com/quick_start/pricing. Prices change, so confirm there before you commit.
Compare DeepSeek API with alternatives
Other llm apis: OpenAI API, Anthropic Claude API, Google Gemini API, Mistral La Plateforme, xAI Grok API.
Looking at quality rather than price? See the DeepSeek API score on APIbenchmarks. Full entry: DeepSeek API pricing.