DeepSeek V4 Pro pricing & benchmarks
Released Apr 24, 2026 · served by 18 providers · ranked #21 by intelligence in our index
What DeepSeek V4 Pro is
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...
- Model ID:
deepseek/deepseek-v4-pro - Modality: text->text
- Tool calling: supported · Structured output: supported
- Weights:
deepseek-ai/DeepSeek-V4-Proon Hugging Face
Providers and prices for DeepSeek V4 Pro
The same model costs different amounts depending on who serves it. Prices are per 1M tokens, cheapest first. Uptime and throughput are OpenRouter's rolling measurements, not vendor claims.
| Provider | Input | Output | Context | Quant | Uptime 24h | Throughput |
|---|---|---|---|---|---|---|
| DeepSeekcheapest | $0.435 | $0.870 | 1.0M | — | 99.9% | — |
| Baidu | $0.625 | $1.25 | 1.0M | fp8 | 96.3% | — |
| StreamLake | $0.670 | $1.34 | 1.0M | fp8 | 95.6% | — |
| GMICloud | $0.679 | $1.36 | 1.0M | fp8 | 93.1% | — |
| Ionstream | $1.13 | $2.26 | 1.0M | fp4 | 56.9% | — |
| Novita | $1.17 | $2.34 | 1.0M | fp8 | 99.1% | — |
| Wafer | $1.20 | $2.40 | 1.0M | fp4 | 90.5% | — |
| DeepInfra | $1.30 | $2.60 | 1.0M | fp4 | 99.1% | — |
| DigitalOcean | $1.39 | $2.78 | 262K | — | 90.9% | — |
| Alibaba | $1.42 | $2.83 | 1M | — | 93.2% | — |
| SiliconFlow | $1.50 | $3.13 | 1.0M | fp8 | 97.6% | — |
| Venice | $1.65 | $3.30 | 1M | — | 85.5% | — |
| AtlasCloud | $1.68 | $3.38 | 1.0M | fp4 | 98.5% | — |
| BaseTen | $1.74 | $3.48 | 262K | fp4 | 99.3% | — |
| Parasail | $1.74 | $3.48 | 1.0M | fp8 | 86.4% | — |
| Together | $1.74 | $3.48 | 512K | — | 94.5% | — |
| CoreWeave | $1.74 | $3.48 | 1.0M | fp8 | 97.4% | — |
| Fireworks | $1.74 | $3.48 | 1.0M | — | 87.0% | — |
DeepSeek V4 Pro benchmark results
Design Arena head-to-head results, as reported through the OpenRouter model API.
| Category | Arena | Rank | Elo | Win rate |
|---|---|---|---|---|
| 3d | models | #10 | 1320 | 60% |
| gamedev | models | #21 | 1291 | 55.7% |
| codecategories | models | #24 | 1275 | 54.3% |
| asciiart | models | #27 | 1188 | 46.6% |
| godotgamedev | agents | #27 | 1059 | 34% |
| uicomponent | models | #30 | 1261 | 52.1% |
| website | models | #30 | 1260 | 52.6% |
| fullstack | agents | #33 | 948 | 22.1% |
| webapps | agents | #34 | 1002 | 26.6% |
| svg | models | #36 | 1186 | 46.5% |
| dataviz | models | #39 | 1226 | 49.7% |
Cheaper models in the same class
Models scoring within 4 points of DeepSeek V4 Pro on the Intelligence Index, but with a lower output price.
Compare DeepSeek V4 Pro head-to-head
FAQ
How much does the DeepSeek V4 Pro API cost?
$0.435 per 1M input tokens and $0.870 per 1M output tokens. A typical workload of 1M input + 300K output tokens costs about $0.70.
Is DeepSeek V4 Pro free?
No. It is a paid model starting at $0.435 per 1M input tokens, though some providers offer trial credits.
Which provider is cheapest for DeepSeek V4 Pro?
DeepSeek at $0.435 input / $0.870 output per 1M tokens.
Can I self-host DeepSeek V4 Pro?
Yes — weights are published as deepseek-ai/DeepSeek-V4-Pro on Hugging Face, so you can run it on your own hardware.