Skip to main content
The full platform model catalog for the AI & Models guide. Every Digital Twin has access to the same providers and models listed here; which model is selected for each Model Use is configured per Digital Twin under Personalization.

​
Platform Models

Praxis AI middleware offers access to a broad catalog of state-of-the-art AI models. You can select the model that best fits your needs based on performance, cost, and capabilities. — Default — in a model picker follows the platform default, which Praxis AI keeps on a current, capable model. Keep it unless you need a specific provider, cost or speed. In the pickers, the platform default for a job is marked — Default, and deprecated models are hidden unless you tick Show Deprecated.

​
Provider-Based Models

Praxis AI exposes conversation and related capabilities (vision, audio, embeddings, moderation, realtime) through multiple provider types:
  • Amazon Bedrock
  • OpenAI-Compatible Clients (OpenAI, Cohere, Kimi, OpenRouter)
  • Anthropic Direct API
  • Google Gemini Native SDK (Google AI Studio, or Google AI Platform)
  • Mistral AI Native SDK
  • xAI Native API
  • Stability AI Native API
Each provider contains groups and individual models with specific capabilities and uses. The Typical Uses column names the jobs a model can be picked for. Some are background jobs whose model Praxis AI chooses: KAG (building the knowledge graph from your files), Praxis Shield (spotting misuse in conversations), Meeting roles (helper steps for meeting bots) and Web roles (helper steps around a chat answer). Most conversation models can also run the subagent roles; each role’s picker offers only the models suited to it. On Amazon Bedrock, Claude Opus 5 and Opus 5.5 can be picked for every agent role, the conductor included. The embedding models are listed for reference: the Embeddings Generation picker offers only text-embedding-3-small.
Anthropic models via Bedrock are platform models of choice, mainly for Conversation and Image Analysis. Claude Sonnet 5 and Claude Sonnet 4.6 support the optional 1M token context window (see Inference Settings). In the Thinking column, Adaptive means the model decides how much to think for each request.
Claude Fable 5 / 5.1 on Bedrock — data-sharing opt-in required. These two models refuse to run on Bedrock until your AWS account opts in to Anthropic’s 30-day data retention — and there is no console setting for it yet. Anthropic requires 30-day data retention for Fable/Mythos-class traffic on Bedrock. Your AWS account must set its Bedrock data-retention mode to provider_data_sharing (via the Bedrock Data Retention API — no console UI at launch) before global.anthropic.claude-fable-5 or global.anthropic.claude-fable-5-1 can be invoked; otherwise requests fail with “data retention mode ‘default’ is not available for this model.” Each Fable version also needs its own Bedrock model-access agreement accepted (Bedrock console → Model access). If that data-sharing posture isn’t acceptable for your deployment, use claude-fable-5-1 / claude-fable-5 on the Anthropic Direct API instead, which does not require the Bedrock opt-in.
Deprecated models are on their way out. A Digital Twin still set to one is moved to a current model automatically — normally the platform default. Assistants keep the model they name until you change it, so move them to a newer model.
Stability AI models are no longer available through Bedrock. They are now served via the Stability AI Native API — see the dedicated accordion below.
These models are configured against the OpenAI API and used across Conversation, Image Analysis, Summary, Audio, TTS, Moderation, and Realtime.

​
Conversation / Vision / Summary

​
Image Generation

OpenAI no longer serves DALL·E and GPT Image 1: a request that still names one (for example from an assistant) is answered by GPT Image 2.5 Flare. GPT Image 1.5 and GPT Image 1 mini are deprecated and will follow. Choose GPT Image 2.5 Flare (fast everyday images) or GPT Image 2.5 Sunburst (precise edits) instead.When asked, Pria can produce these images in shapes beyond the default square: square (1024×1024), landscape (1536×1024) and portrait (1024×1536). If a requested size isn’t supported by the chosen model, Pria automatically uses the closest available size.

​
Video Generation

​
Embeddings

​
Audio Transcription and Translation

​
Text-to-Speech (TTS)

​
Moderation

​
Real-Time Speech-to-Speech (RT / STS)

​
Full-Duplex Voice (GPT-Live)

Picked from the OpenAI GPT-Live provider. GPT-Live listens and speaks at the same time; your Digital Twin’s own conversation model, files and tools supply the answers. See Realtime Voice & Avatars.OpenAI Voices (Convo, speech-to-speech): Cedar (New), Marin (New), Alloy, Ash, Ballad, Coral, Echo, Sage, Shimmer, VerseOpenAI Voices (Read-aloud): the voices a Text-to-Speech model can speak depend on the model. tts-1 (the platform default) and tts-1-hd accept Alloy, Ash, Coral, Echo, Fable, Nova, Onyx, Sage and Shimmer; gpt-4o-mini-tts also accepts Ballad and Verse. Cedar and Marin are Convo-only. The Read-aloud voice picker only offers the voices your speech model supports; if a twin was saved with a voice its model cannot speak, read-aloud plays Alloy and the picker says so until you choose another voice.
More information: https://platform.openai.com/docs/models
Mistral AI models are accessed through Mistral’s own API and are used for Conversation, Image Analysis, Summary, Audio, TTS, Embeddings, and Moderation. Requires an API key.

​
Conversation / Vision / Summary

​
Retired by Mistral

Mistral has switched these models off. Any Digital Twin, assistant or platform default that used one is moved automatically to its current equivalent — Mistral Medium 3.5 (mistral-medium-2604) for most, Mistral Small 4 (mistral-small-2603) for the small ones. Where the new model isn’t offered for that particular job, the next current model that is takes its place: Mistral Large 3 (mistral-large-2512) for a summary setting, and Medium 3.5 for an agent-role seat. Custom AI Model definitions are never changed automatically.

​
Audio Transcription (STT)

​
Text-to-Speech (TTS)

​
Embeddings

​
Moderation

More information: https://docs.mistral.ai/getting-started/models/models_overview
xAI models are accessed through xAI’s native API and are used for Conversation, Image Analysis, Summary, Code, Image Generation, Embeddings, TTS, and Real-Time Voice. Requires an API key.

​
Conversation / Vision / Summary

​
Image Generation

These models are shaped by aspect ratio rather than exact pixel size — Pria can request a square, landscape, or portrait shape (e.g. 1:1, 16:9, 9:16, 3:2) and defaults to a square. An unsupported shape is mapped to the closest available aspect.

​
Embeddings

​
Text-to-Speech (TTS)

xAI Voices: Eve, Ara, Rex, Sal, Leo

​
Real-Time Speech-to-Speech (xAI Voice Agent)

xAI RT Voices: Eve (Default), Ara, Rex, Sal, Leo
Audio transcription (STT) for xAI delegates to the configured OpenAI transcription model (e.g., gpt-4o-mini-transcribe).
More information: https://docs.x.ai/docs/models
Stability AI models are accessed through Stability’s own API and are dedicated to media generation: Image and Audio (its video model is retired, see below). Praxis AI provides access by default; to use your own Stability account, add its key under Stability AI API Key in the Digital Twin’s configuration.

​
Image Generation

These models are shaped by aspect ratio rather than exact pixel size — Pria can request shapes such as 1:1, 16:9, 9:16, 21:9, 3:2, or 4:5 and defaults to a square. An unsupported shape is mapped to the closest available aspect.

​
Audio Generation

​
Video Generation

Stability AI retired its video generation API on 2025-07-24. The model is kept in the dropdown for backward compatibility but calls return a deprecation message. Videos are made by Veo 3.1 Fast (Google AI Platform, the default) or Veo 3.1.
Stability AI remains a dedicated media-generation provider for Image and Audio — it does not expose Conversation, Embeddings, STT, or RT Voice. Conversation models from OpenAI, Anthropic, Gemini, Mistral, xAI, or Bedrock hand image and audio requests to Stability when a Stability model is selected for Image Generation or Audio Generation. Videos are made by the model selected for Video Generation (Veo 3.1 Fast by default). Stability models create a new image from the description; they do not edit an existing image.
More information: https://platform.stability.ai/docs/api-reference
  • AI & Models — model uses, selection, reasoning effort, moderation, and BYOM
  • Personalization — where each Digital Twin selects its per-capability models