Skip to content

models.md

Latest commit

 

History

History
203 lines (167 loc) · 7.5 KB

models.md

File metadata and controls

203 lines (167 loc) · 7.5 KB

HelpingAI Models Overview 🤖

This guide provides comprehensive information about all HelpingAI models, their capabilities, and when to use each one.

🌟 Model Lineup

Dhanishta 2.0 Preview - Revolutionary Reasoning AI

Model ID: Dhanishtha-2.0-preview

The world's first AI model with intermediate thinking capabilities. Dhanishta 2.0 can pause, reflect, and restart its reasoning process multiple times within a single response.

🧠 Key Features

  • Intermediate Thinking: Multiple <think>...</think> blocks throughout responses
  • Self-Correction: Identifies and fixes logical inconsistencies mid-response
  • Multi-Phase Reasoning: Complex problems solved through iterative thinking
  • Structured Emotional Reasoning (SER): <ser>...</ser> blocks for empathetic responses
  • Multilingual Capabilities: 39+ languages with consistent reasoning quality
  • Transparent Process: See exactly how the AI arrives at conclusions

📊 Specifications

  • Base Model: Qwen3-14B
  • Parameters: 14.8B
  • Context Length: 40,960 tokens
  • Languages: 39+ (including English, Hindi, Chinese, Spanish, French, German, Japanese, Korean, Arabic)
  • Training: 16.3 days on 8x NVIDIA H100 GPUs

🎯 Best Use Cases

  • Complex mathematical problems
  • Logic puzzles and riddles
  • Educational content requiring step-by-step explanations
  • Multi-step reasoning tasks
  • Research and analysis requiring multiple perspectives
  • Programming problems with detailed solutions
  • Scientific problem-solving

📈 Benchmark Performance

Benchmark Score Type
MMLU 78.1% 1-shot
HumanEval 75.0% 1-shot
ARC 76.0% 1-shot
HellaSwag 81.0% 1-shot
TruthfulQA MC1 75.0% 1-shot
Math 500 95.68% few-shot
AIME 2024 82.81% few-shot

HelpingAI3-raw - Advanced Emotional Intelligence

Model ID: Helpingai3-raw

An advanced language model with enhanced emotional intelligence and contextual awareness, specifically designed for empathetic and supportive interactions.

💝 Key Features

  • Enhanced Emotional Understanding: Deep comprehension of human emotions
  • Contextual Awareness: Understands nuanced social and emotional contexts
  • Therapeutic Responses: Trained on therapeutic conversation patterns
  • Crisis Response: Specialized training for supportive crisis interactions
  • Cultural Sensitivity: Awareness of diverse cultural contexts
  • Empathetic Communication: Natural, caring response style

📊 Training Data

  • 15M Emotional Dialogues: Comprehensive emotional conversation training
  • 3M Therapeutic Exchanges: Professional therapy session patterns
  • 250K Cultural Conversations: Diverse cultural interaction training
  • 1M Crisis Response Scenarios: Emergency and crisis support training

🎯 Best Use Cases

  • Emotional support and counseling
  • Mental health assistance
  • Personal advice and guidance
  • Therapeutic conversations
  • Crisis intervention support
  • Empathetic customer service
  • Educational mentoring
  • Creative writing with emotional depth

🔄 Model Comparison

Feature Dhanishta 2.0 HelpingAI3-raw
Primary Strength Reasoning & Problem Solving Emotional Intelligence
Thinking Process Visible with <think> blocks Internal processing
Self-Correction ✅ Real-time error fixing ❌ Single-pass responses
Multilingual ✅ 39+ languages ✅ Multiple languages
Mathematical Reasoning ⭐⭐⭐⭐⭐ Excellent ⭐⭐⭐ Good
Emotional Understanding ⭐⭐⭐ Good ⭐⭐⭐⭐⭐ Excellent
Creative Writing ⭐⭐⭐⭐ Very Good ⭐⭐⭐⭐⭐ Excellent
Educational Content ⭐⭐⭐⭐⭐ Excellent ⭐⭐⭐⭐ Very Good
Therapeutic Use ⭐⭐⭐ Good ⭐⭐⭐⭐⭐ Excellent

🎯 Choosing the Right Model

Use Dhanishta 2.0 When You Need:

✅ Complex Problem Solving

# Perfect for multi-step reasoning
response = hai.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[{"role": "user", "content": "Solve this step by step: If a train travels 120 km in 2 hours, and then 180 km in 3 hours, what's the average speed?"}],
    hide_think=False
)

✅ Educational Explanations

# Great for teaching concepts
response = hai.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[{"role": "user", "content": "Explain photosynthesis step by step"}]
)

✅ Logic and Reasoning Tasks

# Excellent for puzzles and logic
response = hai.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[{"role": "user", "content": "Three friends have different colored hats. If Alice doesn't have red, Bob doesn't have blue, and Charlie doesn't have green, and there are only red, blue, and green hats, who has which color?"}]
)

Use HelpingAI3-raw When You Need:

✅ Emotional Support

# Perfect for empathetic responses
response = hai.chat.completions.create(
    model="Helpingai3-raw",
    messages=[{"role": "user", "content": "I'm feeling overwhelmed with work and personal life. I don't know how to cope."}]
)

✅ Creative Writing

# Excellent for emotional storytelling
response = hai.chat.completions.create(
    model="Helpingai3-raw",
    messages=[{"role": "user", "content": "Write a heartwarming story about friendship"}]
)

✅ Personal Advice

# Great for life guidance
response = hai.chat.completions.create(
    model="Helpingai3-raw",
    messages=[{"role": "user", "content": "How can I improve my relationships with my family?"}]
)

🔧 Model-Specific Parameters

Dhanishta 2.0 Special Parameters

response = hai.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[{"role": "user", "content": "Your question here"}],
    hide_think=False,  # Show/hide thinking process
    temperature=0.7,   # Lower for logical tasks, higher for creative
    max_tokens=2048    # Allow space for thinking blocks
)

HelpingAI3-raw Optimal Settings

response = hai.chat.completions.create(
    model="Helpingai3-raw",
    messages=[{"role": "user", "content": "Your question here"}],
    temperature=0.8,   # Higher for more empathetic responses
    top_p=0.9,        # Good balance for emotional nuance
    max_tokens=1000   # Sufficient for detailed emotional responses
)

🌍 Language Support

Dhanishta 2.0 Supported Languages (39+)

  • European: English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Polish, Czech, etc.
  • Asian: Chinese (Simplified/Traditional), Japanese, Korean, Hindi, Arabic, Thai, Vietnamese, etc.
  • Others: Hebrew, Turkish, Greek, Swedish, Norwegian, Finnish, and many more

Language-Specific Reasoning

# Reasoning in multiple languages
response = hai.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[{"role": "user", "content": "Explique-moi la photosynthèse étape par étape"}],  # French
    hide_think=False
)

📚 Next Steps

Choose the right model for your use case and unlock the full potential of AI! 🚀