Model Hub

The Model Hub lists the models available from your connections and sets the model Andromity uses. Open it from Settings → Model Hub, or with /settings in the chat. To switch model quickly without opening Settings, use /model in chat or the model control under the prompt box.

Choosing a modelCopy Anchor Link

Select Select Model on a card. Andromity saves the model and its connection as the default in config.toml:

[default]
provider = "openrouter"
model = "deepseek/deepseek-v4.1-flash"

The Current Active banner at the top shows the selected model with its provider, context window and price.

Pick a model that supports tool calling. Andromity works by calling tools to read files, edit and run commands, so a model without tool support can chat but can't complete tasks.

Finding modelsCopy Anchor Link

• Search matches the model name, ID, provider and description, for example sonnet, deepseek or qwen.
• Category chips filter to Pinned, Free Tier, Coding & Agentic, Reasoning / Thinking, Vision & Multimodal and Flagships. Categories other than Pinned and Free Tier come from tags, which only OpenRouter models and some custom endpoints have.
• Provider chips show one provider at a time: Andromity, OpenRouter, Anthropic, OpenAI, Gemini, DeepSeek, Groq, Ollama (Local), NVIDIA NIM, and one chip per custom provider.

The count next to Model Hub in the Settings sidebar is the number of models listed.

PinningCopy Anchor Link

Select the pin on a card to add it to Pinned, your quick list. Pins are saved in config.toml:

[models]
pinned = [
  { id = "auto", provider = "andromity", name = "Andromity Auto" },
  { id = "anthropic/claude-3.7-sonnet", provider = "openrouter", name = "Claude 3.7 Sonnet" },
]

The two pins above are the defaults until you change the list.

Model cardsCopy Anchor Link

Each card shows:

• the model name and exact ID
• the provider
• a description
• the context window, for example 128K ctx or 1M ctx
• Free Tier, Coding and Reasoning badges where they apply
• the price per million input and output tokens, Free Tier, or Pricing unavailable

Using an OpenRouter model that isn't listedCopy Anchor Link

Paste an OpenRouter model ID into the Custom: bar and select Use Custom Model, for example anthropic/claude-3.7-sonnet:thinking. This always uses the openrouter connection, so it needs an OpenRouter key.

For models on your own server, add a custom provider instead.

OllamaCopy Anchor Link

Select the Ollama (Local) chip to see the models installed on your machine. The Hub checks whether Ollama is running and offers the next step:

Ollama models are listed only while Ollama is running. See Local AI Coding with Ollama.

Where the list comes fromCopy Anchor Link

Refresh Catalog asks each connected provider for its current models:

The result is cached in model_live_catalog_cache.json next to config.toml. Until a provider has been refreshed, the Hub shows Andromity's built-in list of models for it.

Context windows come from what the provider reports, then the built-in list, then a default of 131,072 tokens. Prices come from LiteLLM's price list, then OpenRouter's published prices, then the built-in list. The context window decides when Andromity compacts a conversation and what the context indicator in chat shows.

Per-role models for subagentsCopy Anchor Link

The Hub sets the main model. To give a subagent role its own model, use the terminal app:

/subagent                                  # list roles with their model, provider and tools
/subagent set <role> <model> [provider]    # give a role its own model

Role models are stored under [subagents] in config.toml. A role without its own model uses [default] model.