DeepSeek V4 Pro

deepseek-v4-pro by DeepSeek Released April 24, 2026
ReasoningFunction callingStructured outputPrompt cachingWeb searchComputer useImage inputFile inputAudio inputVideo inputEU hostingFree endpointUp to 35% off
Input from
$0.66 / 1M
Output from
$1.98 / 1M
Cached input from
$0.022 / 1M
Context window
1.05M
Endpoints
6
from 6 providers
Regions
Monde, Asie, Europe, États-Unis

About DeepSeek V4 Pro

Un modèle phare de type MoE à grande échelle, comptant 1 600 milliards de paramètres et 49 milliards de paramètres activés, prenant en charge en natif des longueurs de contexte pouvant atteindre un million de tokens. Entraîné sur un vaste corpus de données de haute qualité, il excelle dans le raisonnement mathématique avancé, l'inférence logique complexe, la programmation spécialisée et l'analyse approfondie de textes longs, ce qui le rend particulièrement adapté aux applications exigeantes telles que la recherche de pointe, les flux de travail bureautiques sophistiqués et les agents d'IA avancés.

DeepSeek V4 Pro est un modèle conversationnel développé par DeepSeek, disponible sur Eden AI via 6 endpoints chez 6 fournisseurs, à partir de 0,66 $ par million de tokens en entrée et 1,98 $ par million de tokens en sortie, avec une fenêtre de contexte allant jusqu'à 1,05M tokens.

Providers and prices

6 endpoints from 6 providers, cheapest first, in USD per million tokens. Every one is called with the same Eden AI key: pin one with its endpoint ID, or let Eden AI route for you. .

Endpoint
Region
Context
Input / 1M
Output / 1M
Cached / 1M
Supports
Tencent Cloud (Asie)Official−%tencent/deepseek-v4-pro-202606
Asie
1M
$0.66$0.66
$1.98
$0.022
ReasoningToolsJSONCacheVisionWeb
DeepSeek (Asie)Official−%deepseek/deepseek-v4-pro
Asie
1M
$0.66$0.66
$1.98
$0.022
ReasoningToolsJSONCacheVisionWeb
DeepInfra (États-Unis)Official−%deepinfra/deepseek-ai/DeepSeek-V4-Pro
États-Unis
1.05M
$1.3$1.3
$2.6
$0.1
ReasoningToolsJSONCacheVisionWeb
QwenOfficial−35%qwen/deepseek-v4-pro
Monde
1M
$2.4$1.56
$3.12
$0.13
ReasoningToolsJSONCacheVisionWeb
TensorX (Europe)Official−%tensorx/deepseek/deepseek-v4-pro
Europe
1.05M
$1.75$1.75
$3.5
$0.4375
ReasoningToolsJSONCacheVisionWeb
NebiusOfficial−%nebius/deepseek-ai/DeepSeek-V4-Pro
Monde
1.05M
$1.75$1.75
$3.5
$1.75
ReasoningToolsJSONCacheVisionWeb

Performance

Measured on real Eden AI traffic over the last 30 days. Endpoints with too little traffic are not shown.

Fastest first token
0.84 s
Nebius
Lowest latency
2.2 s
TensorX (Europe)
Highest throughput
115.52 tok/s
TensorX (Europe)
Best availability
100%
Nebius
Endpoint
First token
Latency p50
Latency p95
Throughput
Availability
Cache hits
Tool errors
JSON errors
Tencent Cloud (Asie)
s
s
s
tok/s
%
%
%
%
DeepSeek (Asie)
1.51 s
6.81 s
61.64 s
62.42 tok/s
99.89%
97.27%
0.56%
26.47%
DeepInfra (États-Unis)
2.36 s
5.85 s
45.23 s
35.46 tok/s
99.94%
89.03%
1.63%
0%
Qwen
3.27 s
12.21 s
77.73 s
40.99 tok/s
100%
92.72%
4.78%
27.38%
TensorX (Europe)
1.07 s
2.2 s
21.31 s
115.52 tok/s
99.93%
88.41%
3.29%
0%
Nebius
0.84 s
2.52 s
11.93 s
84.87 tok/s
100%
90.84%
4.89%
0%

Estimate your monthly cost

Enter your expected volume to compare every endpoint at once.

Call DeepSeek V4 Pro with Eden AI

Use the model ID to let Eden AI pick an endpoint, or an endpoint ID from the table above to pin one.

import requests

response = requests.post(
    "https://api.edenai.run/v3/chat/completions",
    headers={"Authorization": "Bearer YOUR_EDENAI_API_KEY"},
    json={
        "model": "",
        "messages": [{"role": "user", "content": "Hello!"}],
    },
)
print(response.json())

Questions about DeepSeek V4 Pro

On Eden AI, prices for this model start at $0.66 per million input tokens and $1.98 per million output tokens. Prices differ by endpoint; the providers table lists each one. .

Up to 1.05M tokens, depending on the endpoint.

6 endpoints from 6 providers, available in: Monde, Asie, Europe, États-Unis. All of them are called with the same Eden AI API key.

Yes. At least one endpoint is hosted in Europe; pin it with its endpoint ID from the providers table to keep requests in the EU.

Over the last 30 days, the fastest first token was 0.84 s (Nebius) and the highest throughput was 115.52 tokens per second (TensorX (Europe)).