DeepSeek API pricing explained (with real cost examples)

DeepSeek API pricing is per million tokens, with separate prices for the tokens you send (input) and the tokens the model writes (output). Here's how it works, with examples using Solo Router's current prices.

Current DeepSeek prices on Solo Router

Model	Input / 1M	Output / 1M	Cached input / 1M
DeepSeek V4.1 Flash	$0.39	$1.56	$0.008
DeepSeek V4 Flash	$0.18	$0.36	$0.036
DeepSeek V4 Pro	$2.26	$4.52	$0.45
DeepSeek V4 Pro (0813)	$1.72	$5.15	$0.057
DeepSeek V3.2	$0.65	$2.08	$0.20

How a request is charged

cost = input tokens × input price + output tokens × output price

• Tokens are pieces of words. As a rough guide, 1,000 tokens is about 750 English words.
• Input includes everything you send: the system prompt, the chat history and the new message.
• Output is what the model writes back.
• Reasoning tokens (the model's "thinking") count as output.
• Cached input: when you resend the same long prefix (like a big system prompt), the repeated part is billed at the much lower cached price.

Example costs

A chat message. About 2,000 input tokens (history plus question) and 400 output tokens on DeepSeek V4.1 Flash costs well under a tenth of a cent. A thousand such messages costs about a dollar.

A coding-agent step. Agents often send 30,000–60,000 tokens of code context per step. Caching matters most here: the repeated context is billed at the cached price, which is a small fraction of the normal input price.

Every Solo Router response includes usage.cost, the exact price of that request in USD, so you never have to estimate.

Tips to keep DeepSeek costs down

1. Use V4.1 Flash or V4 Flash for routine work and save V4 Pro for hard problems.
2. Keep system prompts stable so they get cached.
3. Trim old chat history instead of resending everything forever.
4. Set max_tokens so a runaway answer can't cost more than you expect.
5. Give each app its own key with a spending limit.

See all prices on the DeepSeek API page or compare every model on the Models page.