Over the last few days, I’ve spent some time playing around making different Bots (BoodleBox’ equivalent of Custom GPTs). If you plan on free Bbox users using one of your Bots, which do you choose? This blog entry is going to be my reference guide for the future.

The Problem

One of the screens you see often is this one when creating a Bot:

But what’s the difference between Premium and Free models in BoodleBox?

When you mouse over over the question marks in the circles, you’ll see:

For free BoodleBox users, this means realistically, you can only create Bots that are based on the free models because otherwise they will only get five uses. I wish they allowed 20 uses…that would work in a workshop. And, of course, the users must have a Boodle Box account…I wish if we used the free models, users wouldn’t need an account. Sigh. I have a similar complaint about ChatGPT.

Comparing the Boodle Box Models

One of the questions that goes around in my head is, “What is the best model to use, when?” I asked Google Gemini 2.5 Pro to come up with a long and short comparison.

Here’s the short comparison:

Premium Models

• Claude 3.7 & 3.5 Sonnet: Extremely fast for how smart they are, making them ideal for enterprise-scale tasks and interactive work like coding or editing.
• ChatGPT 4o: The best for real-time, human-like voice conversations and interacting with the world through video and audio.
• Pi: The most “human-like” and empathetic AI, designed for supportive and engaging conversations rather than complex tasks.
• Cohere: Built for businesses, focusing on privacy and using your own company data to provide accurate, grounded answers.

Free Models

• ChatGPT 4o-mini: A faster, cheaper version of GPT-4o, great for powering everyday apps and chatbots at a large scale.
• Claude 3.5 Haiku: The fastest model of the bunch, perfect for applications needing instant responses like live customer support.
• Gemini 1.5 Pro: The best for analyzing enormous amounts of information at once, like entire books, long videos, or massive codebases.

---

More detailed info below this notice of free BoodleBox classes:

Premium Models

These models represent the cutting edge of AI, offering the most advanced capabilities, often with a subscription fee or pay-per-use pricing. They are geared towards users and businesses that require the highest levels of performance, reasoning, and multimodal understanding.

Model	Best Features	What It’s Best Used For
Claude 3.7 Sonnet	Next-Generation Reasoning: As a likely successor to the 3.5 series, it is expected to feature even more advanced reasoning, interpretation, and code generation capabilities. Enhanced Speed & Performance: Aims to provide intelligence approaching top-tier models like Opus but at a significantly faster speed and lower cost. Sophisticated Vision Analysis: Built to interpret complex charts, graphs, and visual data with high accuracy.	* Handling complex, multi-step workflows that require a high degree of logical reasoning. * Demanding enterprise tasks like financial forecasting from reports and analyzing intricate datasets. * Serving as a powerful, high-speed “engine” for complex AI applications.
Claude 3.5 Sonnet	Exceptional Speed-to-Performance Ratio: Operates at twice the speed of Anthropic’s top-tier Claude 3 Opus, making it incredibly fast for its high level of intelligence. “Artifacts” Feature: A unique capability that allows users to edit and expand on AI-generated content (like code or documents) in a separate window, enabling a more interactive workflow. Top-Tier Vision Capabilities: Excels at tasks requiring visual reasoning, such as accurately transcribing text from imperfect images or interpreting complex charts.	* Enterprise-scale AI deployments, such as sophisticated customer support bots and data interpretation dashboards. * Interactive coding, where a developer and AI can iterate on code in real-time. * Workflows that involve creating and refining content or designs based on AI output.
ChatGPT 4o	True Multimodality: Seamlessly understands and generates a combination of text, audio, and images. Its ability to hold a real-time voice conversation with tone and emotion detection is revolutionary. Speed and Accessibility: Offers GPT-4 level intelligence but is significantly faster and more widely available, even to free users (with limits). Real-time Interaction: Can “watch” a screen or camera feed and respond to what it sees, opening up possibilities for live tutoring, presentations, and accessibility tools.	* Natural, real-time voice assistance and dynamic, interactive conversations.<br> * Live translation and interpretation services.<br> * Creative content generation that blends different media types (e.g., creating a story with accompanying images).
Pi	High Emotional Intelligence (EQ): Designed from the ground up to be an empathetic, supportive, and engaging conversational partner. It excels at natural, human-like dialogue. Conversational Focus: Its primary strength is in dialogue. It remembers context and personal details to create a tailored and coherent conversational experience. Concise and Helpful Nature: Tends to provide clear, to-the-point responses, making it useful as a sounding board or personal assistant.	* Brainstorming, personal coaching, and providing emotional support. * Users seeking a more collaborative and less transactional AI interaction. * Practicing conversations, preparing for interviews, or simply having an engaging chat.
Cohere	Enterprise and Privacy Focus: Built for business applications with a strong emphasis on data privacy. It can be deployed in a virtual private cloud or on-premise. Retrieval-Augmented Generation (RAG): Specializes in “grounded generation,” using a company’s own documents and data to provide accurate, verifiable answers and reduce hallucinations. High Customization: Offers powerful tools for businesses to fine-tune models on their specific data, creating a bespoke AI for their unique needs.	* Building internal knowledge base search tools and enterprise-grade chatbots. * Developing scalable AI products that require high data security and accuracy. * Powering customer support systems that rely on specific company policies and product information.

---

Free Models

These models offer remarkable capabilities at no cost, making powerful AI accessible to everyone. They are often slightly smaller or have different performance characteristics than their premium siblings but are perfect for everyday tasks, development, and experimentation.

Model	Best Features	What It’s Best Used For
ChatGPT 4o-mini	Excellent Balance of Cost, Speed, and Intelligence: Intended to be a highly capable yet much cheaper and faster alternative to the full GPT-4o, making it ideal for high-volume applications. Strong Multimodal Capabilities: Retains the ability to understand and process text, images, and other inputs effectively. High Speed for Real-Time Use: Optimized for low-latency responses, crucial for powering responsive chatbots and other interactive tools.	* Task automation and content classification at scale. * Powering mobile applications and consumer-facing chatbots that require quick, intelligent replies. * Serving as a cost-effective backend for various AI-powered features in software products.
Claude 3.5 Haiku	Blazing Speed: The fastest model in the Claude 3.5 family, designed for near-instantaneous responses, making it feel highly responsive in a chat interface. Compact and Cost-Effective: Its small size and efficiency make it one of the most affordable models on the market for its capability class. Robust Safety Features: Like all Anthropic models, it has strong safeguards against harmful or inappropriate outputs.	* Live customer support where immediate answers are critical. * Content moderation to quickly filter large volumes of user-generated content. * Simple, high-volume tasks like data extraction or real-time translation.
Gemini 1.5 Pro	Massive Context Window: Features a groundbreaking 1 million token context window (expandable to 2 million), allowing it to analyze enormous amounts of information at once (e.g., a 1,500-page PDF or hours of video). Native Multimodality: Was built from the start to understand and reason across text, code, images, audio, and even video files seamlessly. “Needle in a Haystack” Retrieval: Excels at finding and recalling specific facts or details from within vast, unstructured datasets.	* Analyzing entire codebases for bugs or documentation. * Summarizing and answering detailed questions about long documents, academic papers, or books. * Processing and analyzing the content of video and audio files without needing a transcript.