DeepSeek API pricing
DeepSeek sells API access to its V4 models at aggressive prices. Input is priced by cache hit vs miss — hits are ~30x cheaper — and everything is half price during off-peak hours.
Official price list: api-docs.deepseek.com
Current lineup
How DeepSeek billing works
Input is billed by prompt-cache outcome: a cache MISS bills the full input rate, a cache HIT bills ~1/30th of it. There is no separate cache-write fee.
Time-of-day pricing: peak hours are 01:00–04:00 and 06:00–10:00 UTC; every other hour is off-peak at exactly half the peak rate — on input, cached input, and output alike.
Thinking and non-thinking modes of the same model cost the same per token; both V4 models take 1M tokens of context and up to 384K output.
The models are open-weight, so hosts (Fireworks, Together) sell them too at flat per-token prices — Fireworks' V4 Flash undercuts DeepSeek's own peak rate, while DeepSeek's cache hits and off-peak windows undercut every host. Compare against your traffic pattern.
Every DeepSeek model we price
The full catalog — including dated snapshots and variants — exactly as Marginal prices incoming usage events.
| Cached input / 1M | ||||
|---|---|---|---|---|
| deepseek-chat | $0.28 | $0.42 | 128K | $0.028 |
| deepseek-chat | $0.28 | $0.42 | 128K | $0.028 |
| deepseek-coder | $0.14 | $0.28 | 125K | — |
| deepseek-r1 | $0.55 | $2.19 | 64K | — |
| deepseek-reasoner | $0.28 | $0.42 | 128K | $0.028 |
| deepseek-reasoner | $0.28 | $0.42 | 128K | $0.028 |
| deepseek-v3 | $0.27 | $1.10 | 64K | $0.07 |
| deepseek-v3.2 | $0.28 | $0.40 | 160K | — |
| deepseek-v4-flash | $0.44 | $1.32 | 1M | $0.014 |
| deepseek-v4-flash | $0.14 | $0.28 | 1M | $0.0028 |
| deepseek-v4-pro | $1.32 | $3.96 | 1M | $0.044 |
| deepseek-v4-pro | $0.435 | $0.87 | 1M | $0.0036 |
Prices last synced Aug 20, 2026 from the LiteLLM community price catalog — the same catalog Marginal prices real usage events against, refreshed daily. Spot an error? Tell us.
Other providers
Using DeepSeek in production?
Marginal prices every call your app makes against this catalog and slices spend by customer, feature, or any field you define. One track() call to integrate.