DeepSeek V3

Model Information

Display Name: DeepSeek V3

API Model ID: deepseek-ai/deepseek-v3

Category: Text To Text

Description: DeepSeek V3 is a groundbreaking Mixture-of-Experts model with 671B total parameters, activating only 37B per token for exceptional efficiency. **Key Features:** - 163K token context window - MoE architecture (671B total, 37B active) - Multi-head Latent Attention - Structured outputs (JSON) - Prompt caching support - Cost-effective inference **Capabilities:** - High-quality text generation - Reasoning and math - Code generation - JSON mode for structured outputs - Long-context processing **Best For:** - Cost-sensitive applications - High-volume inference - Complex reasoning at low cost - Long document processing **Technical Specs:** - Architecture: Mixture-of-Experts - Total Parameters: 671 billion - Active Parameters: 37 billion per token - Efficient inference design

Context Window: 163,840 tokens

Max Output: 16,384 tokens

How to Use This Model

To use DeepSeek V3 via the HInow.ai API, use the model ID: deepseek-ai/deepseek-v3

API Request Example (Chat/Text)


POST https://api.hinow.ai/v1/chat/completions
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{
  "model": "deepseek-ai/deepseek-v3",
  "messages": [
    {"role": "user", "content": "Your message here"}
  ]
}
              

Pricing

  • input: $0.19
  • output: $0.38

Available Parameters

  • temperature: Controls randomness (0-2). Default: 0.7 (Options: 0, 0.3, 0.5, 0.7, 1.0, 1.5, 2.0)
  • top_p: Nucleus sampling (0-1). Default: 0.9 (Options: 0.1, 0.5, 0.7, 0.9, 0.95, 1.0)
  • max_tokens: Max tokens to generate (1-16384) (Options: 512, 1024, 2048, 4096, 8192, 16384)
  • repetition_penalty: Reduce repetition (0.01-5). Default: 1 (Options: 1.0, 1.1, 1.2, 1.5, 2.0)
  • response_format: Output format (Options: text, json_object, json_schema)

Quick Reference

To use this model, set: "model": "deepseek-ai/deepseek-v3"

Featured: No

Documentation: https://hinow.ai/models/deepseek-ai/deepseek-v3

API Endpoint: https://api.hinow.ai/v1

Back to Models
DeepSeek V3

DeepSeek V3

deepseek-ai/deepseek-v3

$0.190 / $0.380
per 1M tokens (in/out)

About

DeepSeek V3 is a groundbreaking Mixture-of-Experts model with 671B total parameters, activating only 37B per token for exceptional efficiency.

Key Features:

  • 163K token context window
  • MoE architecture (671B total, 37B active)
  • Multi-head Latent Attention
  • Structured outputs (JSON)
  • Prompt caching support
  • Cost-effective inference

Capabilities:

  • High-quality text generation
  • Reasoning and math
  • Code generation
  • JSON mode for structured outputs
  • Long-context processing

Best For:

  • Cost-sensitive applications
  • High-volume inference
  • Complex reasoning at low cost
  • Long document processing

Technical Specs:

  • Architecture: Mixture-of-Experts
  • Total Parameters: 671 billion
  • Active Parameters: 37 billion per token
  • Efficient inference design

Capabilities

Text To Text
Context164K tokens
Max Output16K tokens

Parameters

temperature

Controls randomness (0-2). Default: 0.7

00.30.50.71.01.52.0
top_p

Nucleus sampling (0-1). Default: 0.9

0.10.50.70.90.951.0
max_tokens

Max tokens to generate (1-16384)

512102420484096819216384
repetition_penalty

Reduce repetition (0.01-5). Default: 1

1.01.11.21.52.0
response_format

Output format

textjson_objectjson_schema

Code Examples

curl -X POST https://api.hinow.ai/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $HINOW_API_KEY" \
  -d '{
    "model": "deepseek-ai/deepseek-v3",
    "messages": [
      {"role": "user", "content": "Hello! How are you?"}
    ],
    "parameters": {
      "temperature": "0",
      "top_p": "0.1",
      "max_tokens": "512",
      "repetition_penalty": "1.0",
      "response_format": "text"
    }
  }'