DeepSeek V4 Pro 0813 API

Run long agent and coding sessions on the August build of DeepSeek V4 Pro, from $0.522 per million input tokens.

1M context window
1.7T total parameters
FP8 precision served
MIT licensed weights

from openai import OpenAI

client = OpenAI(
    base_url="https://api.modelsheep.com",
    api_key=API_KEY_MODELSHEEP,
)

response = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-V4-Pro-0813",
    messages=[
        {
            "role": "user",
            "content": "Why do sheep count people?"
        }
    ],
)

print(response.choices[0].message.content)

How to use the DeepSeek V4 Pro 0813 API

1
Create a ModelSheep account and copy an API key from the dashboard.
2
Point the OpenAI SDK at the base url https://api.modelsheep.com and set the model to deepseek-ai/DeepSeek-V4-Pro-0813.
3
Call /chat/completions with your messages and with reasoning_effort set to low, high or max, because the model rejects the common medium default, and read the answer from the first choice of the response.

What DeepSeek V4 Pro 0813 is

DeepSeek V4 Pro 0813 is the general release of DeepSeek V4 Pro. The digits name the build date, 13 August 2026. DeepSeek built it on the preview structure, added a DSpark speculative decoding module, and reports higher agent and coding scores than the preview. The undated repository still holds the April preview, so the dated name pins one fixed set of weights.

What it does well

Drives a terminal agent through a long task, one tool call after the next.
Writes and repairs code across a whole repository.
Holds a full document set in one request, up to 1M tokens.

What it is not for

Images and audio. The model reads text only. DeepSeek serves image input on
Self-hosting on one machine. The weights are open, but the model needs a

Calling it

Post to /chat/completions and set the model to deepseek-ai/DeepSeek-V4-Pro-0813. Set reasoning_effort to low, high or max. There is no medium level, and many OpenAI-style clients send that value by default. DeepSeek recommends high for agent and coding work, together with a high output token limit, because high and max write long reasoning traces.

deepseek-ai/DeepSeek-V4-Pro-0813 /chat/completions

SDKs and languages

DeepSeek V4 Pro 0813 answers on an HTTP endpoint, so any language that can post JSON reaches it. These are the ones the box above writes out.

OpenAI SDK: PythonTypeScriptGoJavaC#PHPRust
Anthropic SDK: PythonTypeScriptGoJavaC#PHPRust
HTTP: cURLPythonTypeScriptGoRustJavaC#PHP

Specifications

Author	DeepSeek
Category	language
Accepts	text
Returns	text
Total parameters	1.7T
Active parameters	49B
Score	36.3 on Artificial Analysis intelligence index
Languages	not published
Precision served	FP8
Context	1M tokens
Vector	n/a
Pulls	153K in the last 30 daysWeights pulled from Hugging Face, read on 2026-09-24. It counts the repository, not this API.
License	MIT

Price

Input

$0.522

per million tokens

Output

$1.568

per million tokens

Against OpenRouter

10%

below the cheapest of 18 providers

The cheapest OpenRouter price for this model was $0.581 in and $1.742 out, with a median of $1.320 and $3.960 across the comparable providers. Read on 2026-09-17.

License

DeepSeek publishes the repository and the weights under the MIT license. It permits commercial use, modification and redistribution, so you do not need to read a separate vendor agreement before you ship.

Parameters: 284B, 13B
Context: 1M
Input per 1M: $0.027
Output per 1M: $0.117
Released: Jul 2026
Pulls, 30 days: 4M
License: MIT