Managing models

Every provider connection has its own model list. Open Providers and expand Models on a provider to manage it.

Where the list comes from

• Fetched live. Most providers have a model list endpoint. Max Harness reads it when you connect and whenever you press Refresh models.
• Built-in list. Perplexity and AI21 do not publish a usable list, so a short list ships with the preset. It carries a Built-in list badge because it can fall behind. A few other presets use a built-in list as a fallback if the live request fails.
• Added by you. Type any model ID into the box under the list and press Add model. Use this for new or private models the list does not include. IDs must match the provider's naming exactly.

Refreshing keeps your choices: enabled state, favourites, hand-added models, capability overrides and prices you entered.

Enable, disable and favourite

Large catalogs are easier to use when you hide what you do not need.

• Untick a model to remove it from the model menu.
• Use the filter box with Enable shown or Disable shown to change many at once.
• Star a model to list it first.

Capabilities

Badges show what a model can take:

• Images. Accepts image attachments.
• PDF. Accepts PDF files directly. Without it, PDFs are sent as extracted text.
• Tools. Can call tools such as web search.
• No tools. The model or provider does not support tools, so the tool button is disabled.

Providers report these in different ways, and some do not report them at all. When nothing is reported there is no badge, and Max Harness lets you try. If the provider then rejects the request, its error message is shown.

To correct a capability, press Details on the model and set it to Yes or No. Setting Accepts images to No makes the workspace refuse image attachments for that model before anything is sent.

Prices and cost estimates

Cost is shown only when a price is known:

• OpenRouter publishes per-token prices in its model list, and reports the actual cost of each request. That figure is shown as reported.
• Anthropic prices come from Anthropic's published list as of the date shown in the tooltip.
• You can enter a price for any model under Details, in US dollars per million tokens.

For everything else, replies show token counts without a dollar figure. An estimate multiplies the reported tokens by the list price. It does not account for cached input discounts, batch pricing or tiered rates, so treat it as a guide and check the provider's invoice for the real amount.

Default model

The Default model menu on each provider decides which model new chats start with when that provider is selected. If it is left on "First enabled model", favourites win.