Models / 
DeepSeek
 / 
DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731

Designed for coding-heavy and high-volume workloads that need strong quality at low cost, served on a dedicated endpoint or fine-tuned on your own data.

Overview

Run DeepSeek V4 Flash 0731 on Crusoe. A 284B open-weight model with a 1M-token context, ready to fine-tune and serve with high-performance inference.

Features

Try now
fine_tuning
$6.00
Tailor DeepSeek V4 Flash 0731 to your needs with fully managed fine-tuning. With a few clicks, you can upload your dataset, set your hyperparameters, and ship the trained checkpoint.
self_serve_deployment
Run DeepSeek V4 Flash 0731 on your own dedicated GPUs powered by Crusoe’s optimized serving stack for consistent performance and strong reliability—with no rate limits.
serverless_inference
$0.14 / $0.28 / $0.03
Access DeepSeek V4 Flash 0731 through Crusoe’s serverless API with usage-based, pay-per-token pricing. Connect using our REST endpoints, the Python client, or any OpenAI-compatible SDK.
Model:
deepseek-ai/Deepseek-V4-Flash
Copy model ID
from openai import OpenAI

client = OpenAI(
    base_url='https://api.inference.crusoecloud.com/v1/',
    api_key='<YOUR_API_KEY>',
)

response = client.chat.completions.create(
    model='deepseek-ai/Deepseek-V4-Flash',
    messages=[
        {
            'role': 'user',
            'content': 'Hello, how are you?'
        }
    ],
)

print(response.to_json())
import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://api.inference.crusoecloud.com/v1/',
  apiKey: '',
});

client.chat.completions
  .create({
    model: 'deepseek-ai/Deepseek-V4-Flash',
    messages: [
      {
        "role": "user",
        "content": "Hello, how are you?"
      }
    ],
  })
  .then((response) => console.log(response));
curl 'https://api.inference.crusoecloud.com/v1/chat/completions' \
  -X 'POST' \
  -H 'Authorization: Bearer ' \
  -H 'Content-Type: application/json' \
  -H 'Accept: text/event-stream' \
  -d '{
    "model": "deepseek-ai/Deepseek-V4-Flash",
    "messages": [
      {
        "role": "user",
        "content": "Hello, how are you?"
      }
    ]
  }'  
Copy model ID

Start building with DeepSeek V4 Flash 0731

Metadata

Status
AVAILABLE
Provider
DeepSeek

Specification

Mixture-of-experts
Yes
Parameters
284000000000

Supported
Functionality

Serverless
Supported
Fine-tuning
Supported
Self-serve
Supported
Provisioned throughput
Context length
1048576
Function calling
Supported
Chat
Supported
Image input
Not supported
Voice
Not supported

Start building with DeepSeek V4 Flash 0731

Explore additional models

Thinking Machines Lab
Inkling Small FP8 Dynamic
View model
Thinking Machines Lab
Inkling Small
View model
Reflection AI
Beam 1.0 501B A23B
View model
DeepSeek
DeepSeek V4.1 Flash 1017
View model
Alibaba
Qwen3.6 35B A3B FFT
View model
Alibaba
Qwen3.8 27B FFT
View model
Alibaba
Qwen3.5 9B Base
View model
DeepSeek
DeepSeek V4.1 Flash
View model
NVIDIA
Nemotron 3.5 Super VL 120B A12B
View model
NVIDIA
Nemotron 3 VoiceChat
View model
OpenAI
GPT-OSS 20B
View model
NVIDIA
Nemotron 3 Nano 30B
View model
NVIDIA
Nemotron 3.5 Lightning 30B A3B
View model
Google
Gemma 4 31B
View model
Alibaba
Qwen3.5 4B
View model
OpenAI
GPT-OSS 120B
View model
Meta
Llama 3.3 70B Instruct
View model
Alibaba
Qwen3.6 35B A3B
View model
Alibaba
Qwen3 8B
View model
NVIDIA
Nemotron 3 Nano Omni
View model
Google
Gemma 4 26B A4B
View model
Z.ai
GLM-5.3 Flash
View model
Z.ai
GLM-5.3
View model
Z.ai
GLM-5.2
View model
Alibaba
Qwen3.5 35B A3B
View model
Alibaba
Qwen3.5 2B
View model
NVIDIA
Nemotron 3 Super 120B
View model
Meta
Llama 3.1 8B Instruct
View model
Alibaba
Qwen3.8 27B
View model
Alibaba
Qwen3.6 27B
View model
Alibaba
Qwen3.5 9B
View model
Alibaba
Qwen3 235B A22B Instruct
View model

Frequently
asked questions

Are you ready to build something amazing?