DeepSeek-R1-Distill-Qwen-32B

Preview

DeepSeek-R1-Distill-Qwen-32B is a distilled version of DeepSeek's R1 model, fine-tuned from the Qwen-2.5-32B base model. This model leverages knowledge distillation to retain robust reasoning capabilities. Delivering exceptional performance on mathematical and logical reasoning tasks, it achieves near-o1 level capabilities with faster response times. With its massive 128K context window, native tool use, and JSON mode support, it excels at complex problem-solving while maintaining the reasoning depth of much larger models.

Try now on Groq Card

Key Technical Specifications

Model Architecture

Built upon the Qwen-2.5-32B framework, the model is comprised of 32 billion parameters. The distillation process fine-tunes the base model using outputs from DeepSeek-R1, effectively transferring reasoning patterns.

Performance Metrics

The model demonstrates strong performance across various benchmarks:

AIME 2024: Pass@1 score of 72.6
MATH-500: Pass@1 score of 94.3
CodeForces Rating: Achieved a rating of 1,691

Technical Details

FEATURE	VALUE
Context Window (Tokens)	128K
Max Output Tokens	16,384
Max File Size	-
Token Generation Speed	~140 tps
Input Token Price	$0.69/1M tokens
Output Token Price	$0.69/1M tokens
Tool Use
JSON Mode
Image Support

Use Cases

Mathematical Problem-Solving

Effectively addresses complex mathematical queries, making it valuable for educational tools and research applications.

Coding Assistance

Supports code generation and debugging, beneficial for software development.

Logical Reasoning

Performs tasks requiring structured thinking and deduction, applicable in data analysis and strategic planning.

Best Practices

Prompt Engineering: Set the temperature parameter between 0.5 and 0.7 (ideally 0.6) to prevent repetitive or incoherent outputs.
System Prompt: Avoid adding a system prompt and include all instructions within the user prompt.

Get Started with DeepSeek-R1-Distill-Qwen-32B

Experience the reasoning capabilities of deepseek-r1-distill-qwen-32b with Groq speed now:

shell

pip install groq

Python

from groq import Groq
client = Groq()
completion = client.chat.completions.create(
    model="deepseek-r1-distill-qwen-32b",
    messages=[
        {
            "role": "user",
            "content": "Explain why fast inference is critical for reasoning models"
        }
    ]
)
print(completion.choices[0].message.content)