DeepSeek API pricing: peak, off-peak and cache-hit rates per 1M tokens
What deepseek-flash and deepseek-v4-pro cost per 1M tokens, how DeepSeek’s peak hours and cache hits change the price, and a worked monthly cost example.
Last updated October 22, 2026
How much does the DeepSeek API cost?
On October 22, 2026, deepseek-flash cost $0.30 per 1M input tokens (cache miss) and $1.20 per 1M output tokens at peak times, and deepseek-v4-pro cost $1.32 and $3.96. Off-peak rates are half those prices, and input tokens that hit DeepSeek’s cache cost $0.006 and $0.044 per 1M at peak.
What does each DeepSeek model cost per 1M tokens?
DeepSeek’s models and pricing page lists two chat models. Both have a 1M-token context length and a maximum output of 384K tokens, and both support thinking and non-thinking modes. Input is priced by whether it hits DeepSeek’s cache, and every price has a peak and an off-peak rate.
| Model (version) | Input, cache hit (peak / off-peak) | Input, cache miss (peak / off-peak) | Output (peak / off-peak) |
|---|---|---|---|
| deepseek-flash (DeepSeek-V4.1-Flash-1017) | $0.006 / $0.003 | $0.30 / $0.15 | $1.20 / $0.60 |
| deepseek-v4-pro (DeepSeek-V4-Pro-0813) | $0.044 / $0.022 | $1.32 / $0.66 | $3.96 / $1.98 |
Prices checked October 22, 2026. Providers change prices often; confirm on DeepSeek’s models and pricing page.
Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, excluding Chinese public holidays; all other hours are off-peak. The page lists a concurrency limit of 2,500 for deepseek-flash and 500 for deepseek-v4-pro, and vision support on deepseek-flash only.
How does DeepSeek API billing work?
- Tokens times price. DeepSeek’s deduction rule is simply number of tokens × price, with input and output tokens counted separately.
- Cache hits are cheap. Input tokens that DeepSeek serves from its cache are billed at the cache-hit rate, 2% of the cache-miss price on deepseek-flash and about 3% on deepseek-v4-pro. The page lists no separate charge for writing to the cache.
- The time of day matters. The same request costs half as much off-peak as at peak.
- Prepaid balance. Fees are deducted from your topped-up or granted balance, with granted balance used first.
- Search and Fetch are extra. DeepSeek’s Search and Fetch APIs cost $0.75 per 1,000 successful requests each. Calls made through DeepSeek Harness, Claude Code or Codex are free until per-request billing starts at 00:00 Beijing time on November 1, 2026.
What would 1M requests a month cost on DeepSeek?
Take 1,000,000 requests a month, each with 1,000 input tokens and 300 output tokens: 1,000M input tokens and 300M output tokens. The first four rows assume no cache hits.
Monthly cost = 1,000 × input price + 300 × output price (prices per 1M tokens)
| Model and timing | Calculation | Monthly cost |
|---|---|---|
| deepseek-v4-pro, all peak, cache miss | 1,000 × $1.32 + 300 × $3.96 | $2,508 |
| deepseek-v4-pro, all off-peak, cache miss | 1,000 × $0.66 + 300 × $1.98 | $1,254 |
| deepseek-flash, all peak, cache miss | 1,000 × $0.30 + 300 × $1.20 | $660 |
| deepseek-flash, all off-peak, cache miss | 1,000 × $0.15 + 300 × $0.60 | $330 |
| deepseek-flash, all peak, 800 of 1,000 input tokens cache hits | 200 × $0.30 + 800 × $0.006 + 300 × $1.20 | $424.80 |
Real traffic usually spans both periods, so your bill lands between the peak and off-peak rows. Count your real prompt sizes with the token calculator.
How can you cut your DeepSeek API bill?
- Run batch work off-peak. Off-peak rates are half of peak rates, and weekends are off-peak all day. Jobs that can wait, such as backfills and evaluations, are good candidates.
- Keep prompt prefixes stable. Cache-hit input costs a small fraction of cache-miss input, so keep system prompts and shared context identical and at the start of each request.
- Use the smaller model where it is good enough. deepseek-flash costs less than a quarter of deepseek-v4-pro per input token and less than a third per output token.
- Compare before you commit. DeepSeek’s low prices come with the trade-offs of any provider: concurrency limits, peak-hour pricing and data-handling terms you should review for your use case.
To estimate what response caching and model routing could save on your traffic, try the LLM gateway savings calculator.
How does Perimattic LLM Gateway help with DeepSeek costs?
Perimattic LLM Gateway sits between your application and model providers. Its features:
- One OpenAI-compatible API in front of DeepSeek, OpenAI, Anthropic, Google, Mistral and open models.
- Routing rules that send each request to a model by cost, latency or task, with automatic fallback to another provider when DeepSeek errors or rate-limits.
- Exact-match and semantic caching of responses.
- Budgets and rate limits per team, API key or customer, with cost tracking and alerts.
Get started: Sign up and our team will walk you through Perimattic LLM Gateway for your workloads.
Where do these DeepSeek prices come from?
All prices on this page come from DeepSeek’s models and pricing page. They are in US dollars, before tax. DeepSeek says product prices may vary and that it reserves the right to adjust them. Prices checked October 22, 2026; providers change prices often — confirm on the provider’s page before you rely on them. Perimattic is not affiliated with DeepSeek.
Compare DeepSeek with other providers
DeepSeek API pricing, answered
When are DeepSeek’s off-peak hours?
DeepSeek’s pricing page sets peak hours at 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, excluding Chinese public holidays. Every other hour is off-peak, including weekends and Chinese public holidays in full, and off-peak rates are half the peak rates.
Which model name should I use for DeepSeek’s cheaper model?
Use deepseek-flash. DeepSeek says it is now served by DeepSeek-V4.1-Flash-1017. The older names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but those models are retired: requests are served by DeepSeek-V4.1-Flash-1017 and billed at the Flash price.
Does DeepSeek have a batch API discount?
Not on its pricing page as of October 22, 2026. The page lists peak and off-peak rates and cache-hit pricing, but no batch discount. Scheduling large jobs into off-peak hours is the closest equivalent, at half the peak price.
Want one API for every model, with costs under control?
Sign up to tell us which models and providers you use, and our team will get in touch to set up the LLM Gateway for your stack.
Sign up