Input
$0.119
Output
$0.238
Context
1M
DeepSeek V4 Flash 0731 costs $0.119 per million input tokens and $0.238 per million output tokens. Cached input is billed at $0.0238 per million, 80% less than fresh input.
Specifications
| Lab | DeepSeek |
| Model id | deepseek/deepseek-v4-flash-0731 |
| Input price | $0.119 per million tokens |
| Output price | $0.238 per million tokens |
| Cached input price | $0.0238 per million tokens |
| Context window | 1,048,576 tokens (1M) |
| Maximum output | 943,718 tokens (944K) |
| Tool calling | yes |
| Image input | no |
| Reasoning | yes |
| First listed | 31 July 2026 |
Models it replaced
The feeds list DeepSeek V4 Flash 0731 in place of deepseek/deepseek-v4-flash. Those models' own prices and limits were different; an article that names one of them describes the model it was written about.
DeepSeek's other models
| Model | Input / 1M | Output / 1M | Context | Listed |
|---|---|---|---|---|
| DeepSeek V4.1 Flash 1017 | $0.022 | $0.192 | 1M | 17 October 2026 |
| DeepSeek V4 Flash Vision Exp | $0.2156 | $0.6468 | 1M | 21 August 2026 |
| DeepSeek V4 Pro 0813 | $0.1035 | $4 | 1M | 12 August 2026 |
| DeepSeek V3.2 | $0.26 | $0.38 | 160K | 1 December 2025 |
| DeepSeek V3.2 Exp | $0.27 | $0.41 | 160K | 29 September 2025 |
| DeepSeek V3.1 Terminus | $0.27 | $1 | 160K | 22 September 2025 |
| DeepSeek V3.1 | $0.25 | $0.95 | 160K | 21 August 2025 |
| R1 0528 | $0.50 | $2.15 | 160K | 28 May 2025 |
| DeepSeek V3 0324 | $0.29 | $1.14 | 160K | 24 March 2025 |
| DeepSeek V3 | $0.2574 | $1.0287 | 160K | 26 December 2024 |
Where these numbers come from
Read from OpenRouter's listing of DeepSeek V4 Flash 0731 every 6 hours into the catalog Founden's builders choose their models from. Prices are OpenRouter's listed price per million tokens: for a lab's closed models that is the lab's own list price, and for an open-weight model it is the price of the host OpenRouter routes to by default, which can differ from the lab's own API and from other hosts. US dollars, before batch, volume or regional discounts. Every model.