DeepSeek V4 Flash 0731 API

Run coding agents and tool loops on an open mixture of experts model from $0.027 per million input tokens.

1.3M token context. 304.2B total parameters. 13B active parameters. MIT license.

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-Flash-0731",
    messages=[
        {
            "role": "user",
            "content": "Why do sheep count people?"
        }
    ],
)

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

How to use the DeepSeek V4 Flash 0731 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, pass the key, and set the model to deepseek-ai/DeepSeek-V4-Flash-0731.
3
Send your messages to /chat/completions, set max_tokens high enough to cover the deliberation tokens, and read the answer from choices[0].message.content.

What DeepSeek V4 Flash 0731 is

DeepSeek V4 Flash 0731 is an open mixture of experts language model from DeepSeek. It reads and writes text only. The 0731 build keeps the architecture of the earlier preview and replaces the post-training, which DeepSeek aimed at agent work. DeepSeek reports that this build beats the larger V4 Pro preview on every agentic benchmark it publishes while activating only a small share of its weights per token.

What it does well

Drives a terminal coding agent through a multi-step repair task.
Calls a tool, reads the result, and decides the next call over a long run.
Takes a whole repository in one prompt and answers questions across the files.

What it is not for

Text only. It accepts no image or audio input.
Reasoning never switches off. The lowest effort setting is still a reasoning

Calling it

Post to /chat/completions with the model string deepseek-ai/DeepSeek-V4-Flash-0731. The OpenAI SDK reaches it after you change the base URL and the key. Set reasoning_effort to low, high or max. Raise the output token limit as you raise the effort because the model spends its first tokens on deliberation and a low ceiling cuts the answer off before it starts.

deepseek-ai/DeepSeek-V4-Flash-0731 /chat/completions

SDKs and languages

DeepSeek V4 Flash 0731 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	304.2B
Active parameters	13B
Score	34.5 on Artificial Analysis intelligence index
Languages	not published
Precision served	FP8
Context	1.3M tokens
Vector	n/a
License	MIT

Price

Input

$0.027

per million tokens

Output

$0.117

per million tokens

Against OpenRouter

10%

below the cheapest of 23 providers

The cheapest OpenRouter price for this model was $0.030 in and $0.130 out, with a median of $0.175 and $0.400 across the comparable providers. Read on 2026-09-17.

License

MIT. It permits commercial use, modification and redistribution. It requires that the license text stays with the model. No extra vendor terms gate the weights, so you can read the license once and ship.

Similar models

Qwen3.8 27BQwen/Qwen3.8-27B-FP8

Parameters: 27.8B
Context: 1M
Input per 1M: $0.180
Output per 1M: $1.980
License: Apache 2.0

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

Parameters: 1.7T
Context: 1M
Input per 1M: $0.522
Output per 1M: $1.568
License: MIT

DeepSeek V4 Prodeepseek-ai/DeepSeek-V4-Pro

Parameters: 1.6T, 37B
Context: 1M
Input per 1M: $0.939
Output per 1M: $1.879
License: MIT

Start using DeepSeek V4 Flash 0731

Point the OpenAI SDK at api.modelsheep.com. Your prompts stay in Germany, and nothing is retained.

Get API key Read the quickstart