Cogito v2 preview - 70B

This is a hybrid reasoning model. To enable thinking mode, pass the following parameter with your request to the model:

    
"chat_template_kwargs": {"enable_thinking": true}
    

Here's an example cURL request with thinking enabled:

    
curl -X POST "https://api.together.xyz/v1/chat/completions" \
  -H "Authorization: Bearer $TOGETHER_API_KEY" \
  -H "Content-Type: application/json" \
    -d '{
    "model": "deepcogito/cogito-v2-preview-llama-70B",
    "temperature": 0.6,
    "chat_template_kwargs": {"enable_thinking": true},
    "messages": [
      {
        "role": "user",
        "content": "What are some fun things to do in New York?"
      }
    ]
}'
    

Here's an example Python request with thinking enabled:

    
from together import Together

client = Together()

chat_template_kwargs = {"enable_thinking": True}

response = client.chat.completions.create(
  model="deepcogito/cogito-v2-preview-llama-70B",
  extra_body={"chat_template_kwargs": chat_template_kwargs},
  messages=[
    {
      "role": "user",
      "content": "What are some fun things to do in New York?"
    }
  ]
)
print(response.choices[0].message.content)
    

‍
Architecture Overview:
• 70B dense parameter architecture with optimized efficiency
• Strong reasoning capabilities in the Cogito model family
• Advanced policy improvement for both reasoning and non-reasoning modes

Training Methodology:
• Dual-mode training improving both standard and reasoning performance
• Signal-based training for thinking process optimization
• Advanced distillation techniques preventing reasoning meandering

Performance Characteristics:
• Excellent reasoning performance in 70B parameter class
• Efficient inference with optimized reasoning chains
• Strong performance across diverse reasoning benchmarks

Reasoning & Analysis:
• Mathematical problem solving with step-by-step verification
• Complex logical reasoning across multiple domains
• Research and analytical tasks requiring structured thinking

Multimodal Applications:
• Visual reasoning and image analysis through transfer learning
• Document analysis combining text and visual elements
• Educational content creation with mixed media

Research & Development:
• Proof of concept for scalable self-improvement techniques
• Foundation for superintelligence research initiatives
• Open source contribution to reasoning model development

‍