| Spec | DeepSeek V4 Pro | DeepSeek V4 Flash |
|---|---|---|
| Lowest input | $1.30DeepInfra | $0.14StreamLake |
| Lowest output | $2.60DeepInfra | $0.28StreamLake |
| Lowest cached input | $0.0036Inference.net | $0.0028Inference.net |
| Context | 1.05M | 1M |
| Max output | 524K | 384K |
| Providers | 4 | 2 |
| Released | April 24, 2026 | April 24, 2026 |
| Weights | Open | Open |
| Lab | DeepSeek | DeepSeek |
| Model ID | deepseek/deepseek-v4-pro | deepseek/deepseek-v4-flash |
USD per 1M tokens, as of November 5, 2026.
| Provider | DeepSeek V4 Pro | DeepSeek V4 Flash |
|---|---|---|
| Inference.net | $2.10 / $4.20 | $0.23 / $0.62 |
| StreamLake | $1.74 / $3.48 | $0.14 / $0.28 |
| DeepInfra | $1.30 / $2.60 | – |
| Novita AI | $1.60 / $3.20 | – |
Input / output, USD per 1M tokens.
| Model | 1,000 chat replies2K in, 500 out each | 1,000 agent tasks50K in, 5K out each | Monthly batch job500M in, 50M out |
|---|---|---|---|
| DeepSeek V4 ProDeepInfra | $3.90 | $78.00 | $780.00 |
| DeepSeek V4 FlashStreamLake | $0.42 | $8.40 | $84.00 |
On each model's cheapest provider by input plus output list price.
Input: DeepSeek V4 Flash at $0.14 per million input tokens vs $1.30 (DeepSeek V4 Pro), 9.3x lower. Output: DeepSeek V4 Flash at $0.28 per million output tokens vs $2.60 (DeepSeek V4 Pro), 9.3x lower. Lowest listed prices as of November 5, 2026.
DeepSeek V4 Pro, with 1.05M tokens vs 1M.
2 providers serve both: Inference.net, StreamLake. DeepSeek V4 Pro has 4 providers in total and DeepSeek V4 Flash has 2.
DeepSeek V4 Pro from $0.0036 per million cached input tokens (Inference.net). DeepSeek V4 Flash from $0.0028 (Inference.net).
Yes. Both use the same OpenAI-compatible endpoint, https://api.routerplus.com/v1, and key. Change the model field from "deepseek/deepseek-v4-pro" to "deepseek/deepseek-v4-flash".
