DeepSeek reasoning endpoint when you want chain-of-thought style work at DeepSeek rates. Watch output tokens: reasoning traces can dominate the bill.
Privacy first counting
TokenCALC counts tokens in your browser when an Exact encoding is available. Prompt text does not need to leave your device for local counting.
Pricing snapshot
USD per 1 million tokens. Verify against the provider source before you lock budget.
| Rate | USD |
|---|---|
| Input / 1M | $0.550 |
| Output / 1M | $2.19 |
| Cached input / 1M | $0.140 |
Model details
| Detail | Value |
|---|---|
| Provider | DeepSeek |
| Context window | 128K |
| Tokenizer | Approx |
| Last verified | 2026-08-02 |
| API model id | deepseek-reasoner |
Notes
Reasoning tokens may be billed with output; confirm on DeepSeek docs.