Selecting Between AI models for your chatbots

A guide to decide how you can select the best AI model out of the 10 different AI model selections to power your AI chatbot.

Selecting Between AI models for your chatbots

We support a number of different models provided by multiple AI service providers, namely OpenAI, Anthropic, Minstral, Llama, Cohere, Gemini and many more, each with their own series of models. Each model has its own unique specifications and use cases, and in this guide we will explain how to choose the best model for your use case.

OpenAI Models

1. GPT-4

Attribute	Details
Description	The original GPT-4 model is a large multimodal model that accepts both text and image inputs, providing advanced reasoning and problem-solving capabilities.
Costs	20 messages per user query and chat response
Strengths	Exceptional at complex language tasks, capable of generating coherent and contextually relevant text. It offers high accuracy and is optimized for chat applications.
Use Cases	Suitable for applications requiring deep understanding and generation of text, such as customer support, content creation, and educational tools.

2. GPT-4 Turbo

Attribute	Details
Description	A variant of GPT-4 that is optimized for speed and efficiency while maintaining high performance.
Costs	10 messages per user query and chat response
Image Reading Capabilities	Available
Strengths	Faster response times compared to the standard GPT-4, making it ideal for real-time applications. Supports vision capabilities and function calling.
Weaknesses	More expensive than the GPT-4o models due to higher intelligence and resources required
Use Cases	Best for chatbots and applications needing quick interactions, such as virtual assistants and interactive games.

3. GPT-4o

Attribute	Details
Description	The "omni" model that is multimodal, accepting both text and image inputs, designed for a broader range of tasks.
Costs	5 messages per user query and chat response
Image Reading Capabilities	Available
Strengths	Higher intelligence than previous models, cost-effective, and capable of handling complex tasks efficiently.
Weaknesses	Complexity in use, higher risk of toxic output, potential inconsistencies in response quality
Use Cases	Ideal for applications that require both text and image processing, such as content creation, data analysis, and customer service.

4. GPT-4o Mini

Attribute	Details
Description	A smaller and more affordable version of GPT-4o, optimized for lightweight tasks while still offering advanced capabilities.
Costs	1 messages per user query and chat response
Strengths	Cost-effective, fast, and capable of handling various tasks.
Weaknesses	Reduced performance on complex tasks, limited functionality compared to larger models, susceptibility to errors
Use Cases	Suitable for smaller applications, such as chatbots, virtual assistants, and basic content generation where speed and cost are priorities.

*GPT-4o mini 60k context is an alternative model to the GPT-4mini model provided that provides for better quality answers due to a larger context window included

In Summary:

If you are working on a budget or with limited resources, GPT-4mini is a suitable choice

• If your application doesn’t require extensive context memory or complex function calling, GPT-4o will serve you well

• If you require your AI to have a large context window because you have a long custom instruction for your chatbot, GPT-4 Turbo (40k) is a good choice.

• If you need your chatbot to constantly cite accurate links or for your chatbot to do advanced maths or calculation, GPT-4 or at least the GPT-4 Turbo model is the way to go

Selecting Between Claude AI models

Anthropic has released several models of its Large Language Model series, each with its own unique specifications and use cases. In this document, we will discuss the most optimal use cases for these models.

The difference between Anthropic AI models lies in their capabilities and pricing tiers. Anthropic's new family of AI models, Claude 3, consists of three models: Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku

1. Claude Opus

Attribute	Details
Description	Exceptional performance on highly complex tasks with near-human fluency and understanding
Costs	20 messages per user query and chat response
Strengths	Exceptional performance on highly complex tasks with near-human fluency and understanding
Weaknesses	Most expensive model in the Claude series
Use Cases	Advanced research tool to look at complex research documents. As a critical analysis assistant for a corpus of reports.

2. Claude Sonnet

Attribute	Details
Description	Exceptional performance on highly complex tasks with near-human fluency and understanding
Costs	20 messages per user query and chat response
Strengths	Exceptional performance on highly complex tasks with near-human fluency and understanding
Weaknesses	Most expensive model in the Claude series
Use Cases	Advanced research tool to look at complex research documents. As a critical analysis assistant for a corpus of reports.

3. Claude Haiku

Attribute	Details
Description	Fastest and cheapest model for near-instant responsiveness. Answers simple queries and requests with unmatched speed. Suitable alternative to the GPT-4mini model.
Costs	1 messages per user query and chat response
Strengths	Most cost-effective model in its intelligence category
Weaknesses	Limited to simple queries, may not handle complex tasks as well as Sonnet or Opus. May not be as fluent in handling multilingual chats.
Use Cases	FAQ bots trained on a simple knowledge base that provide quick answers to common questions without needing in-depth analysis or a nuanced understanding of the user’s query.

In summary, Claude is a popular alternative to OpenAI's GPT models, as some prefer the LLM for its friendlier tone. For others, Claude's AI models work better with their knowledge bases than OpenAI's.

If you are working on a budget or with limited resources, Claude Haiku is a suitable choice

• If your application requires extensive context memory, try the Claude Haiku 60K model.

• If you require your AI to have a large context window because you have a long custom instruction for your chatbot, Claude Sonnet (5k) is a good choice.

• If you need your chatbot to constantly cite accurate links or for your chatbot to do advanced maths or calculation, Claude Opus or at least the Claude Sonnet model is the way to go

Other AI models: Mistral, Llama, Gemini and Cohere AI models

1. Mistral Small

Attribute	Details
Description	Fastest and cheapest model for many simple tasks.
Costs	1 messages per user query and chat response
Strengths	Fastest and cheapest model for many simple tasks. Suitable alternative to the GPT-4mini and Claude Haiku models. EU based AI model.
Weaknesses	Limited to simple queries. May not have the same level of support as the more established models. Performance varies based on task complexity.
Use Cases	FAQ bots trained on a simple knowledge base that provide quick answers to common questions without needing in-depth analysis or a nuanced understanding of the user’s query.

2. Gemini 1.5 Flash

Attribute	Details
Description	AI model designed by Google with real time information access.
Costs	5 messages per user query and chat response
Strengths	Can access up to date information from the web, however may have limited functionality and adherence to base prompts as compared to OpenAI and Claude models
Weaknesses	Queries are not within scope. May not have the same level of support as the more established models. Performance varies based on task complexity.
Use Cases	For research purposes, where the chatbot does not have to adhere to a corporate brand but is allowed freedom to access the web.

3. Command R Plus

Attribute	Details
Description	A model optimized for fast responses, designed for real-time applications.
Costs	5 messages per user query and chat response
Strengths	Fast response times suitable for immediate interactions.
Weaknesses	May sacrifice depth and detail in responses for speed.
Use Cases	Chatbots that require quick, straightforward answers to user inquiries.

4. Llama 7B

Attribute	Details
Description	An open-source model that is lightweight and customizable
Costs	1 message per user query and chat response
Strengths	A model with a large context window, allowing it to have decent quality responses at a relatively low cost
Weaknesses	It may not be able to follow instructions from the base prompt as accurately as Claude and OpenAI models. It may also perform more poorly in multilingual tasks.
Use Cases	Chatbots that require quick, straightforward answers to user inquiries.

5. GPT-Router

Attribute	Details
Description	The GPT router helps to route your queries to the best OpenAI model available to answer your user query
Costs	5 message per user query and chat response
Strengths	The smart router picks the best model available for your user query, allowing you to potentially save on the cost of using a more expensive model.
Weaknesses	Chat response quality may be inconsistent, as the router may not always route to the smartest model.
Use Cases	Chatbots that do not strictly need to stick to their corporate brand or base prompts, used as a simple virtual assistant such as a chatbot embedded within a. small scale help center.

We will be continually updating this page as we introduce more and more AI models. If you would like to request for a custom AI model to be made available on Wonderchat, you can reach out to provide us feedback here.

---

If you have any more questions, feel free to reach out to us at [e-mail]