Lyra AI models and analytics

This page covers AI model selection for Lyra coaches and the analytics data available for monitoring coach usage.

On this page: AI model selection | Analytics

AI model selection

Lyra coaches are powered by large language models (LLMs) — the AI technology that enables conversational, context-aware responses. When configuring a coach, you can select which AI model it uses. Different models have different strengths, and your choice will affect the coach's response quality, speed, and behaviour.

Available models

Model

GPT 4.1

GPT 4o

GPT 5

GPT 5 Thinking

GPT 5 Mini

Claude Sonnet 4

Gemini 2.5 Pro

GPT 5.1

GPT 5.1 Thinking

Claude Sonnet 4.5

GPT 5.2

GPT 5.2 Thinking

Claude Sonnet 4.6

Claude Opus 4.6

How to choose a model

For most coaches, the default model will work well. Consider changing the model if you have specific requirements around response style, speed, or if you require complex thinking.

For general-purpose coaching and knowledge retrieval, a flagship model like GPT-4o or Claude Sonnet will give the most capable, well-rounded responses. These are the best choice for coaches that need to handle a wide range of questions, reason through complex scenarios, or follow nuanced instructions in the Purpose field.

Things to keep in mind:

Model updates happen periodically. AI model providers regularly release improved versions. When a model is updated, your coaches will benefit from improvements in reasoning, accuracy, and speed without requiring any changes to your configuration.

Changing models doesn't change knowledge. The AI model affects how the coach reasons and responds. The information it draws on is determined by your knowledge base. Switching models won't add or remove knowledge; it changes how the coach processes and communicates that knowledge.

Analytics

Lyra captures usage data for each coach, which can be shared with admins on request. This data is currently provided manually and is not yet connected into Fuse Universal Analytics, integration with our platform's native reporting is on the roadmap.

Available data points

Metric

Description

Active users

Unique users who have engaged with any Lyra coach in a given period

Total sessions

Number of coaching sessions started across all coaches

Total threads

Number of individual conversation threads opened

Sessions by coach

Which coaches are getting the most use

Top themes

The most common topics users are asking about across coaches

Messages sent

Total messages sent to a specific coach

Number of threads

Total conversation threads opened with that coach

Average messages per thread

Mean conversation depth — how many exchanges a typical conversation involves

What to look for

  • Active users over total messages. A coach with 50 active users having meaningful conversations is more valuable than one with 5 power users sending hundreds of messages.

  • Low messages per thread. If users are opening threads but only sending one or two messages, they may not be finding value. Review the knowledge base and purpose — the coach may not be answering what users actually need.

  • Top themes vs. knowledge base coverage. If a theme is popular but the coach's knowledge base is thin on that topic, that's a gap worth filling. If users are asking about things the coach isn't designed for, consider whether you need clearer guidance on scope or a new coach for that topic.