Skip to main content

​
About the Provider

DeepSeek is a Chinese artificial intelligence company based in Hangzhou, Zhejiang that focuses on research and development of large language models and advanced AI technologies. The firm emphasizes open innovation in AI, publishing models and research under permissive licenses to make powerful language models widely accessible and support collaborative development in the global AI community.

​
Model Quickstart

This section helps you quickly get started with the deepseek-ai/DeepSeek-R1-0528 model on the Qubrid AI inferencing platform. To use this model, you need:
  • A valid Qubrid API key
  • Access to the Qubrid inference API
  • Basic knowledge of making API requests in your preferred language
Once authenticated with your API key, you can send inference requests to the deepseek-ai/DeepSeek-R1-0528 model and receive responses based on your input prompts. Below are example placeholders showing how the model can be accessed using different programming environments.
You can choose the one that best fits your workflow. 
from openai import OpenAI

# Initialize the OpenAI client with Qubrid base URL
client = OpenAI(
    base_url="https://platform.qubrid.com/v1",
    api_key="QUBRID_API_KEY",
)

# Create a streaming chat completion
stream = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-R1-0528",
    messages=[
      {
        "role": "user",
        "content": "Explain quantum computing in simple terms"
      }
    ],
    max_tokens=16384,
    temperature=0.6,
    top_p=0.95,
    stream=True
)

# If stream = False comment this out
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
print("\n")

# If stream = True comment this out
print(stream.choices[0].message.content)

​
Model Overview

DeepSeek-R1-0528 is the May 2025 update to the original DeepSeek-R1, built on the DeepSeek-V3 backbone with 671B total parameters and 37B active per inference pass (MoE).
  • It achieves 87.5% on AIME 2025 — up from 70% in the previous version — reaching performance on par with OpenAI o1, with reduced hallucinations and enhanced front-end capabilities.
  • This update adds JSON output and function calling support for the first time, making it suitable for structured agentic workflows in addition to advanced reasoning tasks.

​
Model at a Glance

FeatureDetails
Model IDdeepseek-ai/DeepSeek-R1-0528
ProviderDeepSeek
ArchitectureMixture-of-Experts (MoE) Transformer — 671B total / 37B active per token, DeepSeek-V3 backbone
Model Size671B Total / 37B Active
Context Length128K Tokens
Release DateMay 2025
LicenseMIT
Training DataLarge-scale multilingual dataset with reinforcement learning for chain-of-thought reasoning

​
When to use?

You should consider using DeepSeek-R1-0528 if:
  • You need advanced mathematical reasoning and problem solving
  • Your application requires code generation and debugging
  • You are building complex multi-step reasoning pipelines
  • Your use case involves research and in-depth analysis
  • You need JSON-structured output generation
  • Your workflow requires function calling and tool use

​
Inference Parameters

Parameter NameTypeDefaultDescription
StreamingbooleantrueEnable streaming responses for real-time output.
Temperaturenumber0.6Recommended range 0.5–0.7 (0.6 default) to prevent endless repetitions.
Max Tokensnumber16384Maximum number of tokens to generate.
Top Pnumber0.95Nucleus sampling.

​
Key Features

  • 87.5% AIME 2025: Up +17.5% from the previous DeepSeek-R1, matching OpenAI o1 on mathematical reasoning.
  • JSON Output & Function Calling: Newly added structured output and tool use support for agentic workflows.
  • Reduced Hallucinations: Improved factual accuracy and consistency over the original R1.
  • Chain-of-Thought Reasoning: Visible reasoning traces for interpretable multi-step problem solving.
  • 671B MoE Architecture: Frontier-level reasoning with only 37B parameters active per token.
  • MIT License: Fully open-source with unrestricted commercial use.

​
Summary

DeepSeek-R1-0528 is the latest reasoning model from DeepSeek, delivering OpenAI o1-level performance at open-source scale.
  • It uses a 671B MoE Transformer with 37B active parameters, built on the DeepSeek-V3 backbone with RL-based chain-of-thought training.
  • It scores 87.5% on AIME 2025 and introduces JSON output and function calling for the first time in the R1 series.
  • The model supports 128K context, visible reasoning traces, and a recommended temperature range of 0.5–0.7.
  • Licensed under MIT for full commercial use.