HelpingAI Logo

Models

Learn about HelpingAI's available models, their capabilities, and how to choose the right model for your use case.

Available Models

Dhanishtha-2.0-preview

Our flagship model with revolutionary Chain of Recursive Thoughts capabilities and token efficiency.

Model ID: Dhanishtha-2.0-preview

Key Features

  • 🧠 Chain of Recursive Thoughts: Reasons mid-response unlike traditional approach of reasoning on top and then answering
  • ⚡ Token Efficiency: 5x fewer tokens than Deepseek for similar quality responses, avoids redundant reasoning chains
  • 🚀 Fast Inference: 4x faster inference through optimized processing
  • 🔄 OpenAI Compatibility: Drop-in replacement for existing applications
  • 🛠️ Advanced Tool Calling: Sophisticated function calling capabilities
  • 📡 Streaming Support: Real-time response generation
  • 💭 Transparent Reasoning: Optional visibility with hideThink parameter

Specifications

FeatureSpecification
Context Length32,768 tokens
Training DataUp to April 2024
LanguagesOptimized for English, supports 50+ languages
Input Cost$0.50 per 1M tokens
Output Cost$1.50 per 1M tokens
Max Output4,000 tokens

Best Use Cases

  • Complex Reasoning: Mathematical problems, logical analysis, step-by-step thinking
  • Educational Tools: Tutoring systems that show reasoning process
  • Code Assistance: Programming help with transparent logic
  • Content Generation: Efficient, high-quality content creation
  • Research & Analysis: Transparent reasoning for complex analysis
  • Customer Service: Natural problem-solving interactions

Model Comparison

Dhanishtha-2.0-preview vs Traditional Models

FeatureDhanishtha-2.0-previewGPT-4Deepseek
Chain of Recursive Thoughts✅ Mid-response reasoning❌ Pre-processing only❌ Pre-processing only
Token Efficiency✅ 5x fewer tokens❌ Standard❌ Standard
Inference Speed✅ 4x faster❌ Standard❌ Standard
Context Length32K tokens128K tokens200K tokens
Tool Calling✅ Advanced✅ Standard✅ Standard
Streaming✅ Yes✅ Yes✅ Yes
Cost per Quality Unit✅ Lower❌ Higher❌ Higher
hideThink Parameter✅ Unique❌ Not available❌ Not available

Using Models

Basic Usage

Python (using requests)

python
import requests

url = "https://api.helpingai.co/v1/chat/completions"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}
data = {
    "model": "Dhanishtha-2.0-preview",
    "messages": [
        {"role": "user", "content": "Can you help me solve this step by step?"}
    ]
}

response = requests.post(url, headers=headers, json=data)
print(response.json())

Python (using OpenAI SDK)

python
from openai import OpenAI

client = OpenAI(
    base_url="https://api.helpingai.co/v1",
    api_key="YOUR_API_KEY"
)

response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "Can you help me solve this step by step?"}
    ]
)

print(response.choices[0].message.content)

Python (using HelpingAI SDK)

python
from helpingai import HelpingAI

client = HelpingAI(api_key="YOUR_API_KEY")

response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "Can you help me solve this step by step?"}
    ]
)

print(response.choices[0].message.content)

JavaScript (using axios)

javascript
const axios = require("axios");

const response = await axios.post(
  "https://api.helpingai.co/v1/chat/completions",
  {
    model: "Dhanishtha-2.0-preview",
    messages: [
      {
        role: "user",
        content: "Can you help me solve this step by step?",
      },
    ],
  },
  {
    headers: {
      Authorization: "Bearer YOUR_API_KEY",
      "Content-Type": "application/json",
    },
  }
);

console.log(response.data.choices[0].message.content);

JavaScript (using OpenAI package)

javascript
import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "https://api.helpingai.co/v1",
  apiKey: "YOUR_API_KEY",
});

const response = await openai.chat.completions.create({
  model: "Dhanishtha-2.0-preview",
  messages: [
    {
      role: "user",
      content: "Can you help me solve this step by step?",
    },
  ],
});

console.log(response.choices[0].message.content);

JavaScript (using HelpingAI SDK)

javascript
import { HelpingAI } from "helpingai";

const client = new HelpingAI({
  apiKey: "YOUR_API_KEY",
});

const response = await client.chat.completions.create({
  model: "Dhanishtha-2.0-preview",
  messages: [
    {
      role: "user",
      content: "Can you help me solve this step by step?",
    },
  ],
});

console.log(response.choices[0].message.content);

Model-Specific Features

Token Efficiency

Dhanishtha-2.0-preview uses 5x fewer tokens than Deepseek while maintaining quality:

python
# Efficient reasoning without redundant chains {#efficient-reasoning-without-redundant-chains}
response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "Explain quantum computing in simple terms"}
    ]
)

# Gets high-quality response with fewer tokens {#gets-high-quality-response-with-fewer-tokens}
print(response.choices[0].message.content)
print(f"Tokens used: {response.usage.total_tokens}")

Chain of Recursive Thoughts

See how the model thinks with the hideThink parameter:

python
response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "What's 15 * 24? Show your work."}
    ],
    hideThink=False  # Show reasoning process
)

# Response will include <think>...</think> tags with step-by-step reasoning {#response-will-include-thinkthink-tags-with-step-by-step-reasoning}
print(response.choices[0].message.content)

Advanced Tool Calling

Dhanishtha-2.0-preview excels at intelligent tool usage:

python
tools = [{
    "type": "function",
    "function": {
        "name": "calculate_math",
        "description": "Perform mathematical calculations",
        "parameters": {
            "type": "object",
            "properties": {
                "expression": {"type": "string"},
                "show_steps": {"type": "boolean", "default": True}
            }
        }
    }
}]

response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "Can you calculate 15 * 24 and show the steps?"}
    ],
    tools=tools
)

Model Selection Guide

For Emotional Support Applications

Recommended: Dhanishtha-2.0-preview

Why: Built-in emotional intelligence with 98 EQ score provides natural, empathetic responses that understand and validate human emotions.

Example Use Cases:

  • Mental health chatbots
  • Grief counseling applications
  • Stress management tools
  • Emotional wellness apps

For Complex Reasoning Tasks

Recommended: Dhanishtha-2.0-preview with hideThink=false

Why: Chain of Recursive Thoughts engine shows step-by-step thinking process, making it ideal for educational and analytical applications.

Example Use Cases:

  • Educational tutoring systems
  • Mathematical problem solving
  • Logical analysis tools
  • Research assistance

For Cost-Sensitive Applications

Recommended: Dhanishtha-2.0-preview

Why: 5x more token-efficient than comparable models while maintaining high quality output.

Example Use Cases:

  • High-volume customer service
  • Content generation at scale
  • Automated support systems
  • Large-scale data processing

For Real-Time Applications

Recommended: Dhanishtha-2.0-preview with streaming

Why: 4x faster inference with optimized streaming support for responsive user experiences.

Example Use Cases:

  • Live chat applications
  • Real-time translation
  • Interactive gaming
  • Voice assistants

Model Limitations

Context Length

  • Maximum: 32,768 tokens
  • Recommendation: Monitor token usage for long conversations
  • Best Practice: Implement conversation trimming for extended chats

Language Support

  • Primary: English (optimized)
  • Secondary: 50+ languages supported
  • Note: Emotional intelligence works best in English

Content Policies

Dhanishtha-2.0-preview follows responsible AI guidelines:

  • No harmful or illegal content generation
  • No personal information exposure
  • Appropriate content filtering
  • Ethical use enforcement

Performance Optimization

Token Efficiency Tips

  1. Use Specific Prompts: Clear, concise prompts reduce token usage
  2. Set Max Tokens: Limit response length with max_tokens
  3. Optimize Temperature: Lower temperature for factual responses
  4. Leverage Caching: Cache responses for repeated queries

Cost Optimization

python
# Efficient approach {#efficient-approach}
response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "Summarize climate change in 50 words"}
    ],
    max_tokens=75,  # Limit response length
    temperature=0.3  # Lower temperature for factual content
)

# Less efficient approach {#less-efficient-approach}
response = client.chat.completions.create(
    model="Dhanishtha-2.0-preview",
    messages=[
        {"role": "user", "content": "Please provide me with a very detailed and comprehensive explanation about climate change including all the scientific details, historical context, current impacts, and future predictions with extensive examples and case studies"}
    ],
    max_tokens=2000,  # Very long response
    temperature=0.8   # Higher creativity (more tokens)
)

Future Models

We're continuously developing new models and capabilities:

  • Dhanishtha-3.0: Next-generation model with enhanced reasoning
  • Specialized Models: Domain-specific models for healthcare, education, etc.
  • Multimodal Support: Vision and audio capabilities
  • Fine-tuning: Custom model training for enterprise customers

Stay updated on new releases:

Next Steps