All LLM prices / DeepSeek
deepseek-v4-pro price per 1M tokens
deepseek-v4-pro costs $1.32 per 1M input tokens and $3.96 per 1M output tokens, and $0.044 per 1M cached input tokens read.
| Per 1M tokens | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
| deepseek-v4-pro | $1.32 | $3.96 | $0.044 | — |
What a call costs
A typical agent step of 20,000 input and 1,500 output tokens costs $0.0323 on deepseek-v4-pro, so 10,000 of them a month is $323.40. Serving the repeated part of the prompt from cache cuts the input side to $0.044 per 1M.
deepseek-v4-pro on other providers
| Listed as | Input | Output | Cache read | Cache write |
|---|---|---|---|---|
| aihubmix/deepseek-v4-pro | $1.69 | $3.38 | $0.1403 | — |
| azure_ai/deepseek-v4-pro | $1.74 | $3.48 | $0.145 | — |
| dashscope/deepseek-v4-pro | $2.40 | $4.80 | $0.20 | — |
| deepseek/deepseek-v4-pro | $1.32 | $3.96 | $0.044 | — |
| fireworks_ai/accounts/fireworks/models/deepseek-v4-pro | $1.20 | $1.20 | $0.60 | — |
| inferenceapis/deepseek-v4-pro | $1.72 | $5.15 | $0.17 | — |
| novita/deepseek/deepseek-v4-pro | $1.60 | $3.20 | $0.135 | — |
| opencode_go/deepseek-v4-pro | $1.32 | $3.96 | $0.044 | — |
| openrouter/deepseek/deepseek-v4-pro | $0.2088 | $0.4176 | $0.0174 | — |
| qwen_ai_platform/deepseek-v4-pro | $2.40 | $4.80 | $0.20 | — |
| qwencloud/deepseek-v4-pro | $2.40 | $4.80 | $0.20 | — |
| tencent/deepseek-v4-pro | $1.74 | $3.48 | $0.145 | — |
Price history
No change to the listed rates since this log began on 2026-10-04.
Price a call with the free API
No key, no account. It uses this same table, and with "usage_shape": "openai" it takes the cached tokens out of prompt_tokens first, so they aren't billed twice.
curl -s https://api.aight.studio/api/cost \
-H 'content-type: application/json' \
-d '{"model": "deepseek-v4-pro", "input_tokens": 20000, "output_tokens": 1500, "cache_read_tokens": 12000}'
Returns cost_usd, a breakdown per token type, the rates per 1M, and when the table last changed. Rates only: GET https://api.aight.studio/api/pricing?models=a,b. Rate-limited per client; an unknown model is a 404, never $0. Full API docs.
pip install aight && aight-collect --report prices your Claude Code, Codex or Gemini CLI history locally with no account.