Router Desk

HomeDeepSeek API price

DeepSeek API price

OpenRouter does not add a markup on these rates. The number on the provider row is what the credit balance loses, plus the fee you already paid when you bought credits.

Updated 2026-09-27

“DeepSeek API price” on OpenRouter is the pass-through rate for a specific slug. Several V4 Pro revisions are listed at once, and their output prices are not close. Pick the slug before you compare a screenshot.

V4 Pro rates on 27 Sep 2026

SlugInput / MOutput / MContext
deepseek/deepseek-v4-pro (0423)$0.348$0.6961,048,576
deepseek/deepseek-v4-pro-0813$0.245$3.501,048,576

Those figures come from GET /api/v1/models that day. The model page also prints a provider table, and a batch variant of 0813 was listed separately. A job that looks cheap on input can be expensive on output if you call 0813 and the model writes a long answer. 0423’s card describes 1.6T total parameters with 49B active, and a hybrid attention design shared with V4 Flash.

What a million tokens costs in practice

On the 0423 slug, one million input tokens and 200,000 output tokens are about $0.35 + $0.14 = $0.49 before the credit-purchase fee. The same shape on 0813 is about $0.25 + $0.70 = $0.95. Cache reads are cheaper than fresh input on both slugs. The 0423 cache-read field was $0.029 per million.

Older DeepSeek prices

V3 0324, R1 0528, and V3.1 Terminus were still on the site with their own provider rows, mostly between $0.24 and $0.70 per million input. They are not the V4 price. Use the matching guide if the slug in your code is one of those.

Provider uptime changes the routed price. Re-read the model page on the day you start a batch.

Questions

How much is DeepSeek V4 Pro on OpenRouter?

On 27 Sep 2026, slug deepseek/deepseek-v4-pro (the 0423 card) was $0.348 per million input tokens and $0.696 per million output tokens. The GA slug deepseek-v4-pro-0813 was $0.245 and $3.50.

Is that the price on DeepSeek’s own API?

It is the price of the provider OpenRouter is routing to. OpenRouter says it does not mark up inference. DeepSeek’s first-party checkout can still differ.

Does the 5.5 percent fee apply per token?

No. The 5.5 percent fee, $0.80 minimum, is charged when you buy credits. Token prices are deducted from the balance after that.

  • DeepSeek V4 Flash price

    DeepSeek V4 Flash 0731 was $0.021 / $0.32 per million tokens on 27 Sep 2026. The 0423 slug was $0.047 / $0.094. Neither was the free router.

  • DeepSeek on OpenRouter

    OpenRouter lists several DeepSeek slugs at once: V4 Pro, V4 Flash, older V3 chats, and R1 0528. The slug in your request picks the price.

  • OpenRouter pricing and credits

    OpenRouter credits are a prepaid dollar balance. Buying credits costs 5.5% ($0.80 minimum). Inference uses each provider’s own token price.

  • Coding models

    Coding-model choice on OpenRouter is price, context, and tool-call reliability. DeepSeek V4, Qwen3.7 Max, GLM 5.2, and Sonnet 5 are the slugs to price first.

Sources