Models and Providers

Every AI feature in Bitment, from Chat to Agents, runs on a model supplied by an AI provider your firm is connected to. This page explains how the model picker works, how a model choice is also a billing choice, and where your firm's available models come from.

Anyone who uses AI in Bitment sees the model picker; understanding it matters most when you operate in more than one firm, or when you want to know why a certain model isn't available to you.

Where to find it

The model picker appears in the composer of every AI surface: AI Chat's new-chat screen and in-conversation picker, the Assistant, and each agent's editor. Your default model is set in Settings › AI, or from the gear icon next to the New Chat model picker (both edit the same setting). What a model costs, and which firm pays, is tracked in the Tokens app.

Key concepts

The unified model picker

Wherever you pick a model, the dropdown is grouped by credential: one section per connected provider, whether it's yours personally or belongs to one of your firms. Picking a model also picks which credential pays for the call. If you operate in more than one firm, this is how the cost lands on the right one: choosing a model from your Firm A credential bills Firm A; choosing one from your personal credential bills you personally.

An AI agent works the same way, except the choice is made once, in the agent's editor, rather than per message: the model you assign an agent determines which firm's credential bills every run that agent does. Changing an agent's model to one served by a different credential transfers the agent to that firm.

Set your default model

  1. Open Settings › AI.
  2. Under Default model for new chats, pick a model.
  3. New conversations start on that model and its billing credential; you can still change the model per chat.

This is the same setting as the gear next to the model picker on Chat's New Chat screen: edit it in either place and both reflect the change.

Which models are available

The models offered in your picker come entirely from the AI provider credentials your firm (or you personally) is connected to; there are no models available by default with no credential behind them. Bigger, more capable models cost meaningfully more per call than smaller ones, so for routine or high-volume work (like automatic email classification), a smaller model is often the better choice. The Tokens app shows exactly what each model has cost.

Local and self-hosted models (Ollama)

A firm can connect its own self-hosted Ollama server, making any model that server runs available in the same unified picker, labeled distinctly as a local model. This is useful for firms that want to keep certain work on infrastructure they control, or that want a free option for high-volume, low-stakes tasks. Local models still show up wherever any other model does: the picker, the default-model setting, and per-agent model assignment all work the same way.

Bring-your-own credentials and who sets them up

Your firm's AI credentials (whether for a first-party provider like Anthropic or AWS Bedrock, an aggregator like OpenRouter, or your own Ollama server) are provisioned and administered by the Bitment team on your firm's behalf, using your firm's own accounts where applicable. As an operator, you don't manage these credentials directly. What you control is which model you use, day to day, and what it costs, which you track in the Tokens app. If your firm needs a new provider connected, a new model enabled, or an existing credential changed, contact your Bitment team.

Personal vs. firm credentials

A credential can be scoped to just you (Personal) or to a whole firm, in which case every team member and every agent belonging to that firm can use the models it offers. This is the same distinction that runs through billing: a Personal credential bills you individually; a firm credential bills the firm. An operator who works across several firms will typically see several credential sections in their model picker: their own personal one, plus one per firm they're active in.

Tips