DeepSeek API pricing: every model, every rate
DeepSeek publishes different rates for peak and off-peak hours, and discounts cache hits heavily. That makes a single headline number misleading, so below is the standard rate for each model.
Full rate card
40 billable models. 0 carry a rate verified against the vendor's own published pricing; the rest are gateway rates only. USD per 1M tokens, verified 2026-09-16.
| Model | Input | Output | Out/In | Official (in / out) |
|---|---|---|---|---|
deepseek-v4-pro-202606 | $4.5 | $13.5 | 3.0× | not verified |
deepseek-v3.1-fast | $4 | $12 | 3.0× | not verified |
deepseek-v3.2-fast | $4 | $12 | 3.0× | not verified |
deepseek-math-v2 | $2 | $8 | 4.0× | not verified |
deepseek-r1-250120 | $2 | $8 | 4.0× | not verified |
deepseek-r1-250528 | $2 | $8 | 4.0× | not verified |
deepseek-r1-h | $2 | $8 | 4.0× | not verified |
deepseek-r1-searching | $2 | $8 | 4.0× | not verified |
deepseek-reasoner | $2 | $8 | 4.0× | not verified |
deepseek-v3-1-terminus | $2 | $6 | 3.0× | not verified |
deepseek-v3-1-think-250821 | $2 | $6 | 3.0× | not verified |
deepseek-v3-fast | $2 | $8 | 4.0× | not verified |
deepseek-v3.1-think | $2 | $6 | 3.0× | not verified |
deepseek-v4-flash-202605 | $1.5 | $4.5 | 3.0× | not verified |
deepseek-chat | $1 | $1.5 | 1.5× | not verified |
deepseek-coder | $1 | $4 | 4.0× | not verified |
deepseek-r1-distill-qwen-32b | $1 | $3 | 3.0× | not verified |
deepseek-v3-250324 | $1 | $4 | 4.0× | not verified |
deepseek-v3.2-speciale | $1 | $1.5 | 1.5× | not verified |
deepseek-v4-pro | $0.66 | $1.98 | 3.0× | not verified |
deepseek-v4-pro-0813 | $0.66 | $1.98 | 3.0× | not verified |
deepseek-r1-distill-llama-70b | $0.5 | $1.5 | 3.0× | not verified |
deepseek-v3.1-n | $0.5 | $2 | 4.0× | not verified |
DeepSeek-R1 | $0.2899 | $1.1596 | 4.0× | not verified |
deepseek-r1-0528 | $0.2899 | $1.1596 | 4.0× | not verified |
deepseek-v3-1 | $0.2899 | $0.8697 | 3.0× | not verified |
deepseek-v3-1-250821 | $0.2899 | $0.8697 | 3.0× | not verified |
deepseek-v3.1 | $0.2899 | $0.8697 | 3.0× | not verified |
deepseek-v3.1-thinking | $0.2899 | $0.8697 | 3.0× | not verified |
deepseek-r1-distill-qwen-7b | $0.25 | $0.5 | 2.0× | not verified |
deepseek-v4-flash | $0.22 | $0.66 | 3.0× | not verified |
deepseek-v4-flash-0731 | $0.22 | $0.66 | 3.0× | not verified |
deepseek-v4.1-flash | $0.15 | $0.6 | 4.0× | not verified |
DeepSeek-V3 | $0.14495 | $0.5798 | 4.0× | not verified |
deepseek-v3-0324 | $0.14495 | $0.5798 | 4.0× | not verified |
deepseek-v3-search | $0.14495 | $0.5798 | 4.0× | not verified |
deepseek-v3.2 | $0.14495 | $0.217425 | 1.5× | not verified |
deepseek-v3.2-thinking | $0.14495 | $0.217425 | 1.5× | not verified |
deepseek-ocr | $0.108 | $0.108 | 1.0× | not verified |
deepseek-ocr1 | $0.108 | $0.108 | 1.0× | not verified |
Output is billed at a multiple of input on most models. The Out/In column is that multiple — it matters more than the input price once output dominates your bill.
Across 40 billable models the rate spans $0.108 (deepseek-ocr) to $4.5 (deepseek-v4-pro-202606) per million input tokens — a 42× spread. Picking the wrong tier is usually the single most expensive mistake here.
What DeepSeek is good at
DeepSeek is the default choice when output volume dominates — R1-class reasoning at a fraction of US frontier pricing.
See all 827 models → or use the cost calculator with your own token split.