AI Models

AI CTRL Secure Chat offers multiple AI models, each with different strengths and capabilities. Understanding when to use which model will help you get better results for different types of work. You can set your preferred default model in the settings, but you can always switch models for specific conversations based on your needs.

Note: AI CTRL Admins can enable and disable specific models in each deployment. Coordinate with your administrator if you need access to a specific model. Expedient regularly curates the list of models and adds to it as new models become available and pass our security requirements.

OpenAI Models

OpenAI GPT-4.1
An enhanced version of GPT-4 with better coding skills and the ability to work with very long documents without losing context.
Best for: analyzing lengthy contracts or reports, software development, precise instruction following

OpenAI GPT-4o
A versatile model that seamlessly works with text, images, and audio—designed for interactive, real-world applications.
Best for: chatbots, multimodal apps, voice assistants, real-time interactions

OpenAI GPT-4o Mini
A compact version that keeps much of GPT-4o's capability while responding faster and costing significantly less.
Best for: budget-friendly chat, quick interactions, high-traffic applications

Anthropic Models

Anthropic Claude 3 Haiku
The fastest and most affordable Claude model, designed for simple conversations and quick tasks where speed matters more than depth.
Best for: rapid responses, basic chat, high-volume applications, simple content drafting

Anthropic Claude 3.7 Sonnet
A balanced model that handles most everyday tasks well without premium pricing—solid reasoning and writing capability at reasonable cost.
Best for: daily work tasks, content creation, customer support, general business writing

Anthropic Claude Sonnet 4
A stronger version of Sonnet with notably improved coding ability and logical reasoning compared to the 3.7 generation.
Best for: software development, detailed analysis, technical writing, complex instructions

Anthropic Claude Sonnet 4.5
The most refined Sonnet model with enhanced accuracy and the ability to handle longer, more autonomous workflows without losing track.
Best for: multi-step projects, advanced coding, sustained research tasks, detailed technical documentation

Google Models

Google Gemini 2.0 Flash
A quick, responsive model built for everyday business tasks with the ability to understand both text and images.
Best for: email drafting, document summaries, quick questions, general productivity

Google Gemini 2.0 Flash Lite
A stripped-down version optimized for cost and speed, trading some intelligence for faster responses and lower bills.
Best for: simple queries, high-volume use cases, budget-conscious applications

Google Gemini 2.5 Flash
An upgraded version with better problem-solving and coding skills while maintaining quick response times.
Best for: coding assistance, technical questions, faster workflows needing moderate intelligence

Google Gemini 2.5 Pro
A more capable model that handles complex reasoning and understands images, video, and text together more effectively.
Best for: engineering support, detailed analysis, working with multiple file types simultaneously

Google Gemini 3.0 Pro
Google's most advanced model with exceptional reasoning ability and massive context memory—can process entire books worth of information at once.
Best for: comprehensive research, analyzing lengthy documents, sophisticated problem-solving, advanced automation

Specialty & Web Search Models

Perplexity Sonar
Built for quickly summarizing current news and finding answers from recent web content—refreshed with up-to-date information.
Best for: news summaries, market monitoring, quick web research, current events

Perplexity Sonar Pro
A more thorough version with deeper analysis capability for business intelligence and comprehensive literature reviews.
Best for: market research, competitive analysis, industry trend reports, detailed reviews

Perplexity Sonar Reasoning (Deepseek R1)
Combines web search with deeper logical thinking for technical troubleshooting and complex documentation lookup.
Best for: debugging technical issues, finding solutions in documentation, complex technical research

Perplexity Sonar Reasoning Pro (Deepseek R1)
The most rigorous version for high-stakes research requiring precision and comprehensive source verification.
Best for: patent searches, regulatory compliance research, legal analysis, technical due diligence

Deep Research: OpenAI o3
A research-focused configuration that conducts exhaustive literature reviews and systematic fact verification.
Best for: academic papers, comprehensive literature synthesis, reproducible research

Deep Research: OpenAI o4-mini
A faster research assistant that maintains careful reasoning while providing quicker exploratory results.
Best for: preliminary research, iterative investigation, prototype analysis

Deep Research: Perplexity Sonar
Configured for thorough investigation with emphasis on finding and citing reliable sources.
Best for: fact-checking, evidence-based reports, due diligence research

HIPAA: OpenAI GPT-4o
A privacy-compliant version specifically configured for healthcare environments handling protected medical information.
Best for: medical documentation, clinical note summaries, HIPAA-compliant healthcare workflows

Tips for Model Selection

• Start with GPT-5 Mini for most tasks - it's well-balanced and handles most business needs effectively

• Switch to reasoning models when you have complex problems that need step-by-step thinking

• Use web access models when you need current information or real-time data

• Try different vendors (OpenAI, Google, Anthropic) to get different perspectives on the same problem

• Consider the specialized models for specific use cases like image creation or HIPAA compliance

Remember, you can always change models mid-conversation by using the model selector at the top of your chat interface.

Advanced Model Comparison Techniques

AI CTRL Secure Chat allows you to add multiple models to a single conversation, giving you side-by-side responses from different AI systems. This powerful feature helps you get better results by comparing different approaches, perspectives, and capabilities.

To add a second model to your conversation locate the current model selected in the upper left corner of the chat window and select the plus sign after the model’s name. Select a new model from the drop down that appears below the current model. Select the minus button after the model name to remove a model from comparison.

When to Use Model Comparison

• Validate and enhance quality. Different models approach problems uniquely; comparing responses reveals blind spots, validates complex analysis, and helps you identify the strongest solution for critical work.

• Match tools to tasks. Testing models side-by-side shows which capabilities fit your specific needs, letting you choose the right tool instead of guessing.

• Combine complementary strengths. Use one model's analytical depth with another's creative output, or merge detailed reasoning with clear communication to get better results than any single model provides.