DeepSeek API Pricing Calculator

Calculate DeepSeek API cost for V4 Flash or V4 Pro using cache-hit input, cache-miss input, output tokens, and monthly request volume. Rates were checked against DeepSeek's official pricing documentation on July 29, 2026.

Calculate DeepSeek cost See official rate inputs

Quick answer · Official direct API rates verified July 29, 2026

DeepSeek API pricing depends on model, cache status, output length, and request volume.

DeepSeek bills cache-hit input, cache-miss input, and generated output at different per-million-token rates. This calculator keeps those categories separate so a repeated prompt prefix is not accidentally priced as fresh input. It estimates model-token charges only; taxes, currency conversion, retries, routing providers, and infrastructure outside the direct DeepSeek API are not included.

Calculate DeepSeek API pricing for your workload

Choose a model, enter tokens per request and monthly request count, then adjust the editable rates if your contract or a future pricing update differs.

New prompt or context tokens not served from cache.
Reusable prefix tokens that DeepSeek reports as cache hits.

Editable price per 1M tokens (USD)

DeepSeek API pricing used by this calculator

The default values below match the official DeepSeek Models & Pricing table checked on July 29, 2026. DeepSeek states that product prices may change, so the inputs remain editable and production budgets should be rechecked before purchase or launch.

ModelCache hit inputCache miss inputOutputContext
DeepSeek V4 Flash$0.0028 / 1M$0.14 / 1M$0.28 / 1M1M tokens
DeepSeek V4 Pro$0.003625 / 1M$0.435 / 1M$0.87 / 1M1M tokens

Verify current values in the official DeepSeek pricing documentation. The official V4 release note also explains the model names, 1M context window, and retirement of the older deepseek-chat and deepseek-reasoner endpoints.

How to use the DeepSeek API pricing calculator

  1. Select V4 Flash or V4 Pro. Flash is the lower-cost default for high-volume workloads; Pro uses a higher rate for workloads that need the stronger model.
  2. Separate cache hits from cache misses. Put reusable prompt-prefix tokens in the cache-hit field only when your API usage data confirms they were cached.
  3. Enter generated output tokens. Reasoning, code, and agent workflows can produce much longer output than simple chat replies.
  4. Enter monthly requests. Use logs or a conservative traffic forecast instead of multiplying one unusually small test request.
  5. Add a planning buffer. Retry traffic, tool loops, failed calls, and output variance can raise the real bill.
Monthly cost = requests × ((cache-miss input ÷ 1M × miss rate) + (cache-hit input ÷ 1M × hit rate) + (output ÷ 1M × output rate))

DeepSeek token price examples

Support assistant on V4 Flash

Use 2,000 cache-miss input tokens, 1,000 cache-hit tokens, 800 output tokens, and 100,000 monthly requests. The calculator separates the low cache-hit rate from fresh input instead of applying one blended price to all prompts.

Agent workflow on V4 Pro

Switch to V4 Pro, increase output tokens for reasoning and tool use, then test a higher request count. This makes output cost visible before a multi-step agent moves into production.

For a broad multi-provider estimate, use the LLM cost calculator. To compare provider rows and capabilities, open the AI model comparison tool.

What this calculator includes and excludes

Included

  • Cache-hit input token cost
  • Cache-miss input token cost
  • Output token cost
  • Per-request and monthly estimates
  • Editable rates for future or account-specific pricing

Not included

  • Third-party router or cloud-provider markups
  • Taxes, exchange rates, prepaid-credit rules, or contract discounts
  • Application hosting, vector databases, observability, or storage
  • Exact token counting from pasted text
  • Guaranteed invoices or future DeepSeek price changes

Choosing between DeepSeek V4 Flash and V4 Pro

Start with V4 Flash when price and throughput dominate the decision, then test V4 Pro on the subset of requests where model quality materially changes the outcome. Do not route every request to the higher-priced model without measuring task success. A production evaluation should compare response quality, output length, latency, retry rate, and the percentage of input that actually receives a cache hit.

The DeepSeek model pricing page remains the better destination for browsing models and provider rows. This calculator is specifically for turning official direct-API token rates into a workload budget. For broader low-cost options, review the cheapest LLM API guide.

DeepSeek API pricing FAQ

DeepSeek charges per token, with separate rates for cache-hit input, cache-miss input, and output. Enter your per-request tokens and monthly request volume above to estimate a workload-specific total.

The API key itself is not the useful budgeting unit. Cost is driven by token usage and the current account billing terms. Check the official DeepSeek pricing and account balance pages before production use.

As verified on July 29, 2026, the official pricing page lists $0.0028 per 1M cache-hit input tokens, $0.14 per 1M cache-miss input tokens, and $0.28 per 1M output tokens for DeepSeek V4 Flash.

As verified on July 29, 2026, the official pricing page lists $0.003625 per 1M cache-hit input tokens, $0.435 per 1M cache-miss input tokens, and $0.87 per 1M output tokens for DeepSeek V4 Pro.

No. This page calculates cost from token counts you already know. Exact text-to-token counting depends on the model tokenizer and message formatting, so pasted-text counting is outside this calculator.

Common causes include more output than expected, retries, tool-call loops, cache misses, third-party provider pricing, taxes, currency conversion, and rate changes after the verification date.