DeepSeek logo

DeepSeek-V3

DeepSeek DeepSeek V3 N/A

DeepSeek V3 model with strong function calling performance (57.23 BCFL accuracy) and moderate latency.

Model Capabilities

Input

Accepted Formats

Text

Input Length

Context Window: Unknown Tokens

Output

Generated Formats

Text

Output Length

Max Output: Not disclosed Tokens

Core Features & Strengths

Function Calling / Tools
Open Source
Function Calling

Specifications & Pricing

Specifications

Model Size: Unknown params
Architecture: N/A

Pricing

Input / 1M tk

Free

Output / 1M tk

Free

Est. cost in USD

Usage Guidance

Prompting Guide

General Tips

  • **Be Specific & Clear:** Detail the desired outcome, format, tone, and constraints.
  • **Provide Context:** Explain the background, your role, or the purpose (e.g., "Act as a helpful architect...").
  • **Specify Format:** Request bullet points, JSON, markdown, code blocks, etc.
  • **Use Examples (Few-Shot):** Provide 1-3 examples of input/output for complex tasks.
  • **Iterate:** Refine prompts based on the model's responses. Don't expect perfection first try.

Function Calling Example

"Book a meeting room for 3 people tomorrow at 2 PM for 1 hour. Find available rooms first."

// Model should trigger 'find_rooms' then 'book_room'