ai-info

DeepSeek

DeepSeek V4 API Cost Calculator

·

The same model, different price tags. Compare what DeepSeek V4 actually costs per month across official and reseller APIs — the cheapest way to run DeepSeek is often not DeepSeek's own portal.

DeepSeek V4 Flash — cheap workhorse

* OpenRouter shows its cheapest host for deepseek-v4-flash-0731 ($0.05/$0.16); actual cost varies by route.
* DeepSeek Official uses the peak/off-peak midpoint. Off-peak: $0.22/$0.66 → total shown ×0.67. Peak: $0.44/$1.32 → ×1.33. See the peak windows in your timezone →

DeepSeek V4 Pro — reasoning / coding

* OpenRouter routes to multiple hosts — cheapest on deepseek-v4-pro-0813 is $0.58/$1.74 (Alibaba); DeepSeek's own endpoint as an OpenRouter host is $0.66/$1.98. Note the undated deepseek-v4-pro ID routes from $0.87/$1.74, above DeepSeek's off-peak rate. Prices shift frequently.
* DeepSeek Official uses the peak/off-peak midpoint. Off-peak: $0.66/$1.98 → total shown ×0.67. Peak: $1.32/$3.96 → ×1.33. See the peak windows in your timezone →

Cache hits are dramatically cheaper at DeepSeek Official ($0.007–$0.044/M) and low at resellers — if your app reuses long prompts, real bills come in far below these estimates. Krater.ai is subscription-based ($20/mo for 1,500 credits, 350+ models), excluded from per-token comparison — worth a look if your usage is spiky.

Every DeepSeek Official price above is a peak/off-peak midpoint — the real rate swings 2× depending on the hour. DeepSeek Off-Peak Pricing Explained shows the peak windows in your timezone, and works out whether routing around them beats switching provider.

Sources

Every rate on this page comes from one of these, checked 2 Sep 2026:

Priced in USD using DeepSeek's international rates. DeepSeek publishes separate yuan pricing for mainland China; that page uses a different currency and is deliberately not quoted here.

Prices checked August 2026 against each provider's official pricing pages; Novita and DeepSeek Official verified against live dashboards.

Related reading

DeepSeek Off-Peak Pricing ExplainedEvery official price above is a peak/off-peak midpoint. See the exact windows in your timezone, and whether routing around them beats switching provider.DeepSeek Long-Context Flakiness, Measured: Phase Sensitivity in Chunked KV-Cache CompressionA ByteDance Seed paper pinpoints why DeepSeek long-context retrieval flakes: chunked KV-cache compression makes accuracy swing up to 40 percentage points with token position. Numbers and fixes.