Fresh paint, wet floors — we're rebuilding CostOfToken into a cross-provider price comparison. Some corners are still under construction; the full version is coming up soon.

DeepSeek API pricing

Every DeepSeek model we track, with input, cached input and output cost in USD per 1M tokens. DeepSeek is among the cheapest capable APIs, with aggressive cache-hit pricing that rewards repeated prompt prefixes.

18 models · updated 2026-09-25 · prices in USD per 1,000,000 tokens

18
Models tracked
$0.030
Cheapest input — DeepSeek V4 Flash Latest
$0.800
Priciest input — R1 Distill Llama 70B
1.3M
Largest context — DeepSeek V4 Flash Latest

Listed in the order DeepSeek publishes them.

DeepSeek models with input, cached input and output pricing per million tokens
DeepSeek Pro LatestProvider API$0.260$0.250$3.501M
DeepSeek Flash LatestProvider API$0.040$0.010$0.4901M
DeepSeek V4.1 FlashProvider API$0.300$0.006$1.201M
DeepSeek V4 Flash Vision ExpProvider API$0.220$0.007$0.6601M
DeepSeek V4 Pro 0813Provider API$0.462$0.015$1.391M
DeepSeek V4 Flash LatestProvider API$0.030$0.016$0.3201.3M
DeepSeek V4 Flash 0731Provider API$0.030$0.016$0.3201.3M
DeepSeek V4 Pro 0423Provider API$0.789$0.066$1.581M
DeepSeek V4 Flash 0423Provider API$0.049$0.010$0.0981M
DeepSeek V3.2Provider API$0.269$0.135$0.400164K
DeepSeek V3.2 ExpProvider API$0.270—$0.410164K
DeepSeek V3.1 TerminusProvider API$0.270$0.135$1.00164K
DeepSeek V3.1Provider API$0.250$0.130$0.950164K
R1 0528Provider API$0.500$0.350$2.15164K
DeepSeek V3 0324Provider API$0.250—$1.00164K
R1 Distill Llama 70BProvider API$0.800—$0.8008K
R1Provider API$0.700—$2.5064K
DeepSeek V3Provider API$0.320—$0.890164K

DeepSeek pricing questions

How much does the DeepSeek API cost?
DeepSeek pricing runs from $0.030 to $0.800 per 1M input tokens depending on the model. R1 Distill Llama 70B costs $0.800 per 1M input and $0.800 per 1M output tokens.
What is the cheapest DeepSeek model?
DeepSeek V4 Flash Latest is the cheapest DeepSeek model we track, at $0.030 per 1M input tokens and $0.320 per 1M output tokens.
Does DeepSeek charge less for cached tokens?
Yes. Cached input is billed at a lower rate than fresh input — for DeepSeek Pro Latest it is $0.250 per 1M versus $0.260, about 4% cheaper. Caching applies to repeated prompt prefixes.
How often is this DeepSeek pricing updated?
Every day. Prices are read directly from the published pricing page, normalized to USD per 1,000,000 tokens, and only recorded when a number actually changes — so the history shows real price movements rather than daily noise.

Compare with other providers