DeepSeek › DeepSeek V3 › Pricing
$0.28 / $0.42per 1M tokens in / out
$0.028cache read · −90%
$0.02per task at non-thinking?
#170on OpenRouter by tokens · 17.8B / week
Price sheetper 1M tokens
| Input | $0.28 | list |
| Output | $0.42 | |
| Cache read | $0.028 | −90% vs input |
| Context · max output | 128K · 8K | tokens |
Where to buy3 providers · per 1M tokens
| Provider | Input | Output | Context |
|---|---|---|---|
| deepseeklab | $0.28 | $0.42 | — |
| DeepInfra | $0.32 | $0.89 | 164K |
| Novita | $0.89 | $0.89 | — |
In DeepSeek's line-upper 1M in / out
| DeepSeek OCR | $0.03 / $0.10 | −85% |
| DeepSeek R1 Distill Llama 8B | $0.20 / $0.32 | −27% |
| DeepSeek R1 Distill Qwen 32B | $0.27 / $0.27 | −14% |
| DeepSeek V3 | $0.28 / $0.42 | this model |
| DeepSeek V4.1 Flash 1017 | $0.30 / $1.20 | +67% |
| DeepSeek R1 Distill Qwen 14B | $0.43 / $1.12 | +92% |
| R1 0528 | $0.39 / $1.63 | 2.2× |
Cheaper for the same workvs DeepSeek V3
| Agnes 2.5 Pro Beta | -52% | score ×4.2 | $0.10 / $0.30 |
| GPT-6 Luna medium | -12% | score ×3.5 | $0.017 |
| GPT-5.6 Luna medium | -21% | score ×2.9 | $0.016 |
| Ling 3.0 Flash VL thinkingOpen | -71% | score ×2.9 | $0.06 / $0.18 |
| MiMo v2 Omni thinking | -44% | score ×2.8 | $0.14 / $0.28 |
Pricing questions
How much does DeepSeek V3 cost?›
$0.28 per 1M input tokens and $0.42 per 1M output tokens. Cached input is $0.028. One Intelligence Index task costs about $0.02 at the default effort.
Is DeepSeek V3 free to use?›
There is no free API tier, but the weights are open: you can run it on your own hardware for the cost of compute.
How much does prompt caching save?›
Cache reads cost $0.028 per 1M, 90% below the input price.
What is a cheaper alternative to DeepSeek V3?›
Agnes 2.5 Pro Beta costs -52% per task with a score ×4.2 on Humanity's Last Exam.