DeepSeek API Pricing
Compare input and output token prices across AI language models.
Provider
Modalities
Context window
Model name
.
Providers:DeepSeek
| Provider | Model | Input per 1M tokens | Output per 1M tokens | |
|---|---|---|---|---|
 DeepSeek | DeepSeek OCR-2Open weights | $0.03 | $0.03 | View |
| DeepSeek | DeepSeek V3.2Open weights | $0.25 | $0.80 | View |
| DeepSeek | DeepSeek V3.1Open weights | $0.27 | $1.00 | View |
| DeepSeek | DeepSeek V3.1 TerminusOpen weights | $0.27 | $1.00 | View |
| DeepSeek | DeepSeek V3Open weights | $0.27 | $1.12 | View |
| DeepSeek | DeepSeek V4.1 FlashOpen weights | $0.30 | $1.20 | View |
| DeepSeek | DeepSeek V3 TurboOpen weights | $0.40 | $1.30 | View |
| DeepSeek | DeepSeek V4 ProOpen weights | $0.45 | $0.89 | View |
| DeepSeek | DeepSeek R1Open weights | $0.70 | $2.50 | View |
| DeepSeek | DeepSeek R1 TurboOpen weights | $0.70 | $2.50 | View |
| DeepSeek | DeepSeek R1 Distill Llama 70BOpen weights | $0.80 | $0.80 | View |
DeepSeek API cost calculator
Enter the tokens of a typical request to compare what it costs with every model in the table. Not sure how many tokens your prompt has? Paste it into the token calculator.
| Model | Per request | Per month |
|---|---|---|
| DeepSeek OCR-2 (open weights, third-party host) | $0.000075 | $0.75 |
| DeepSeek V3.2 (open weights, third-party host) | $0.0009 | $9.00 |
| DeepSeek V3.1 (open weights, third-party host) | $0.001 | $10.40 |
| DeepSeek V3.1 Terminus (open weights, third-party host) | $0.001 | $10.40 |
| DeepSeek V3 (open weights, third-party host) | $0.0011 | $11.00 |
| DeepSeek V4.1 Flash (open weights, third-party host) | $0.0012 | $12.00 |
| DeepSeek V4 Pro (open weights, third-party host) | $0.0013 | $13.45 |
| DeepSeek V3 Turbo (open weights, third-party host) | $0.0014 | $14.50 |
| DeepSeek R1 Distill Llama 70B (open weights, third-party host) | $0.002 | $20.00 |
| DeepSeek R1 (open weights, third-party host) | $0.0027 | $26.50 |
How DeepSeek API pricing works
DeepSeek bills its API per token, with separate prices for input tokens (your prompt) and output tokens (the generated answer). The table shows standard rates per 1 million tokens; a token is roughly three quarters of an English word.