Models
Model identifiers use the form providerID/modelID.
Select A Model
- UI: model picker
- TUI:
/models - CLI flag:
--model provider/model
Default Model Fields
model: primary modelsmall_model: lightweight helper model for low-cost background tasks
{ "model": "<provider>/<model-id>", "small_model": "<provider>/<small-model-id>"}Check the Selected Model
The composer or session model selector shows your active choice. Set model in configuration for a default, or use the picker or CLI flag for the current work. Confirm the selection when opening or resuming a session. A listed model still needs valid provider access.
BYOK And Cost Tiers
AlphaBase is bring-your-own-key (BYOK): you connect provider credentials, and that provider bills you directly for usage. AlphaBase does not mark up or resell model access.
AlphaBase does not guarantee a free model or a bundled usage allowance. Some provider catalogs expose no-cost models, and local runtimes may avoid hosted API charges, but availability, performance, infrastructure cost, and provider terms are configuration-dependent. Use the active model selector and provider documentation as the source of truth. See Providers.
Use Only Local Models
Local Ollama inference does not require a provider API key, ChatGPT login, or Copilot subscription. Install Ollama, keep it running, and download a model that runs on your device. Models tagged :cloud use a hosted service and are not private local models.
Open the composer’s model picker, expand Ollama Local under My Models, and select an installed model. AlphaBase checks an untested model before relying on it. A connected Ollama server means its inventory is accessible; it does not prove that every installed model can run the selected task. Models that fail compatibility checks remain visible with their status rather than disappearing.
Choosing a model switches the composer to manual selection. Keep routing limited to local models if you enable Smart Router. Model inference can work offline after installation; model downloads, web tools, remote connections, and provider-backed voice features still need their respective services.
For complex multi-step or multi-file tasks, validate the model you connect on real work before relying on it as your default. Check the resulting files and tests, rather than relying on a catalog label.
Model Availability
| State | Meaning |
|---|---|
untested | No current usable check is available. |
ready | A recent request or local readiness check succeeded. |
degraded | Access needs rechecking, for example after a rate limit or stale result. |
unavailable | Setup, authentication, or provider availability prevents use. |
These labels describe availability, not a permanent quality ranking. If a request stalls or fails, inspect the error and provider status before retrying.
Model Routing (Opt-In, Advanced)
Routing is optional and must be explicitly enabled for the request. Where supported, select approved candidate models and review the selected model and usage after execution. Routing does not grant provider access or a free allowance.
For SDK integrations, the request’s agentOptions.modelRouter can specify enabled and candidate models. It is not a top-level alphabase.json setting. Confirm compatibility with the API reference supplied for your installed version.
Context Limits
Models can handle a finite amount of text and tool information in a request. If you see a context or token-limit error, read the exact message in Activity, focus the request on relevant files, or start a new conversation with a short summary. Check the selected model’s supported limits and your provider allowance before retrying. A limit error does not grant access to another model.
Per-Model Overrides
provider.<id>.models.<modelID> supports options and variants.
{ "provider": { "<provider-id>": { "models": { "<model-id>": { "options": { "timeout": 300000, }, "variants": { "high": { "disabled": false }, "max": { "disabled": false }, }, }, }, }, },}Variant Notes
- Variants are provider/model-specific.
- You can disable unwanted variants with
variants.<name>.disabled. - Agent-level model + variant settings can override global defaults.
Provider/Model Discovery
Use the model selector, CLI, SDK, or published API contract for capability-aware model discovery. Internal provider route names are not a public compatibility contract.