We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Pricing is quote-based.
DeepSeek API has no free plan listed. You pay from your balance for model usage per 1 million tokens, with rates varying by model, cache status and peak period. Search and Fetch are charged separately per successful request. There is no monthly or annual plan listed. To estimate a project’s bill, count input and output tokens separately, distinguish cache hits from misses, and include weekday peak usage and any Search or Fetch calls.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| deepseek-flash | — | Quote | Developers seeking lower token rates |
| deepseek-v4-pro | — | Quote | Developers needing the Pro model |
Add-ons and overages
- Search API: $0.75 per 1,000 successful requests
- Fetch API: $0.75 per 1,000 successful requests
02What the plan grid doesn't tell you
Cache misses
Flash cache-miss input is $0.15 off-peak or $0.30 at peak per 1M tokens, versus $0.003 or $0.006 for cache hits.
Peak-period rates
For Pro, peak rates apply 01:00–04:00 and 06:00–10:00 UTC on weekdays; listed peak rates are twice off-peak rates.
Search and Fetch
Each API costs $0.75 per 1,000 successful requests, billed separately from model tokens.
Concurrency limits
Flash lists a 2,500 concurrency limit and Pro 500. Confirm applicable rate limits and throttling before projecting throughput.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Side project with routine inference | deepseek-flash | It has the lower listed token rates; estimate cache misses and output tokens, not only cache hits. |
| Application requiring the Pro model | deepseek-v4-pro | This is the listed Pro model; account for its higher token rates and 500 concurrency limit. |
| High-volume Flash workload | deepseek-flash | Its listed concurrency limit is 2,500, but check rate limits and expected cache mix with the vendor. |
04Is DeepSeek API worth the money?
DeepSeek API can be worth paying for when its models fit the workload and token-based billing suits traffic that varies over time. Flash has low listed cache-hit input rates, but cache misses and output tokens cost more, and weekday peak rates are higher. There is no free tier to use as a production allowance, so start with a representative token and request estimate. For fixed budgets, include Search and Fetch calls and verify concurrency and rate-limit behavior before committing.
DeepSeek API pricing FAQ
Is DeepSeek API usage free?
No free plan is listed. Model requests are charged per token and deducted from your balance; Search and Fetch are charged per successful request. There is no monthly or annual plan listed.
How much does DeepSeek API cost per token?
Rates are quoted per 1 million tokens and vary by model, cache status and time. Flash cache-hit input ranges from $0.003 to $0.006, cache-miss input from $0.15 to $0.30, and output from $0.60 to $1.20.
When do DeepSeek API peak rates apply?
Peak rates apply 01:00–04:00 and 06:00–10:00 UTC on weekdays. For deepseek-v4-pro, the listed peak rates are double the off-peak rates. Include the relevant time windows when estimating costs for scheduled workloads.
How much do DeepSeek Search and Fetch cost?
The Search API costs $0.75 per 1,000 successful requests, and the Fetch API also costs $0.75 per 1,000 successful requests. These charges are separate from the per-token costs for model usage.