What DeepSeek is good at
- Very low token prices: deepseek-flash is $0.15 per 1M input and $0.60 per 1M output tokens off-peak, with cache hits from $0.003.
- Both models have a 1M token context and up to 384K output tokens, with tool calls and JSON output.
- The API speaks both the OpenAI and the Anthropic request formats, so existing tools can point at it with a base URL change.
Where DeepSeek breaks
- DeepSeek's privacy policy says it collects, processes and stores personal data in the People's Republic of China and may use it to train its models; you have a right to opt out of training.
- Prices double during peak hours (01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday), so the cheapest rate depends on when jobs run.
- There is no free API tier: usage is deducted from a topped-up balance, concurrency is capped per account (500 for deepseek-v4-pro) with 429 errors above it, and DeepSeek reserves the right to change prices.
When a custom build beats DeepSeek
When data residency matters, running an open-weight model on infrastructure you control, or through a host in your region, keeps the low cost without sending data to DeepSeek's servers.
Setup
Needs a developer. Plan for a developer to set it up and keep it running.
Sources
- api-docs.deepseek.com/quick_start/pricing
- api-docs.deepseek.com/quick_start/rate_limit
- cdn.deepseek.com/policies/en-US/deepseek-privacy-policy.html
- deepseek.com/en
Price checked 9 Oct 2026. If something here is out of date, email indra@kratuva.com and we will fix it.
