Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.praxis-ai.com/llms.txt

Use this file to discover all available pages before exploring further.

Your Digital Twin can talk to many different AI models — fast and cheap ones for quick lookups, deep “thinking” models for research, multimodal ones that read images or generate them. The Conversation Model picker in the sidebar lets you swap models mid-session without losing your conversation.

​
Why switch models

Different jobs deserve different tools.

Cost

Smaller models burn far fewer credits per turn. If you’re brainstorming or doing simple lookups, a lightweight model is plenty.

Speed

Compact models reply in seconds. Reserve the slower, larger ones for when you actually need depth.

Depth

Frontier reasoning models (Claude Opus, GPT-5, Gemini 3 Pro, Grok 4) plan, reflect, and self-check before answering — best for analysis, long planning, and nuanced writing.

Specific capability

Some models accept image input, some generate images, some support live voice, some have million-token context windows. Pick the one whose superpower matches the task.

​
Where to find the picker

1

Open the left sidebar

Click the menu icon (or press the sidebar shortcut) to open the side panel if it isn’t already pinned.
2

Expand Settings

Scroll past Conversations and Favorites to the Settings section and click to expand it.
3

Choose Conversation Model

The first picker in Settings is Conversation Model. The current model is shown; click to open the dropdown and pick a new one.
The change applies to the next question you ask. Answers already on screen stay attributed to whichever model produced them.

​
Platform Default

When you first start chatting, Pria uses the model your Admin has chosen as the default for this Digital Twin — shown as Platform Default in the picker. This is almost always the right starting point — your Admin picked it for a balance of capability, cost, and reliability that fits your institution. Leaving the picker on Platform Default also means you’ll automatically pick up improvements when your Admin upgrades the default to a newer model. You don’t have to chase model releases yourself.

​
Platform models vs Custom Models

Pria ships a curated catalog of models from major providers (Anthropic, OpenAI, Google, AWS Bedrock, xAI, Mistral). These appear in the dropdown by their friendly names — Claude Sonnet 4.7, GPT-5, Gemini 3 Pro, Grok 4, and so on.Capabilities, prices, and context limits are kept up to date by the Praxis AI team. You don’t need to wire up API keys to use them.

​
Per-assistant model override

Assistants (specialised personas in your Digital Expert Gallery) can pin themselves to a specific model. When you launch a chat with such an assistant, the Conversation Model picker shows the assistant’s pinned model and switching is disabled — the assistant author chose that model deliberately because it fits the persona’s task. To use a different model with the same prompt, start a new conversation without the assistant, or with an assistant that doesn’t lock its model.
This is a common pattern for image-generation assistants, code-review assistants, and other specialists where the right model is part of the assistant’s identity.

​
Model capabilities at a glance

The dropdown shows a few icons next to each model so you can see at a glance what it can do:
IconMeaning
Thinking — supports the reasoning effort scale.
Tools — can call file search, web search, image generation, code interpreter, and the other built-in tools.
Image input — you can paste, drop, or upload images into the chat for the model to see.
Image output — can generate images natively (without delegating to a separate image model).
Realtime voice — can power Convo Mode.
Hover any model row to see its full capability summary, max context window, and price tier.

​
What changes when you switch

  • In-flight answers keep their original model. A reply being streamed from GPT-5 stays on GPT-5 even if you switch the picker mid-stream.
  • The very next question uses the new selection.
  • Conversation memory is shared. Your earlier turns, files, and citations all carry over — only the brain answering the new question changes.
  • Thinking display, knowledge mode, and other settings carry over unchanged. They’re attached to the conversation, not the model.
If you switch from a model with thinking to one without (or vice versa), the reasoning effort picker may grey out or change defaults. You can flip it back any time.

​
When to use Default vs. explicit selection

  • You don’t have a strong reason to change. Your Admin’s pick is almost always a sensible balance.
  • You want automatic upgrades as your institution adopts newer models.
  • You’re chatting with an assistant that already pins its own model.
  • You need a specific capability (image input, image generation, live voice, long context).
  • You’re chasing speed for short factual lookups → pick a smaller model.
  • You’re doing deep research or long-form writing → pick a frontier model and turn up reasoning effort.
  • You’re comparing answers across models to validate a tricky claim.
  • Reasoning & Thinking — how to control how deeply each model thinks before answering.
  • Knowledge Modes — how the model uses your uploaded files.
  • Convo Mode — live voice conversation with the realtime models.
  • Assistants — building and using model-pinned personas.
  • Credits — how model choice affects credit spend.