aura models

List available models from the configured provider.

Syntax

aura models [flags]

Description

Lists all available models from the configured provider with optional sorting. Model visibility can be controlled per-provider using models.include and models.exclude glob patterns in the provider config.

Model lists are cached per provider connection in .aura/cache/models/. Metadata learned when Aura inspects a selected model is saved in .aura/cache/model-metadata/ and reused by listings, including /model, after the model is unloaded. Fresh explicit capability reports take precedence over saved values. Listing models does not load or swap workers to discover capabilities.

T means thinking is supported and V means vision is supported. T? and V? mean the provider has not reported that capability; a blank means it is known to be unsupported. Models with unreported tool support appear under Tool capability unknown, rather than Other. Capability filters include only confirmed support.

Use --no-cache to bypass both catalog and metadata caches and re-fetch from providers (fresh data is still written back to cache). Unknown capabilities can therefore reappear with --no-cache. Use aura cache clean to delete all cached data.

Flags

Flag	Short	Default	Description
--sort-by		size	Sort by: name (alphabetical), context (context window length), size (model file size on disk)
--filter	-f		Filter by capability (combinable): thinking, tools, embedding, reranking, vision
--name	-n		Filter by model name (wildcard patterns, repeatable)

Plus all global flags.

Examples

# List models from default provider (Ollama)
aura models

# Sort by context window size
aura models --sort-by context

# List models from specific providers
aura --providers openrouter models

# List models from multiple providers
aura --providers ollama,openrouter models

# Filter by capability
aura models --filter vision
aura models --filter thinking,vision

# Filter by model name (wildcard)
aura models --name='gemma*'
aura models --name='qwen*' --name='llama*'

# Combine name and capability filters
aura models --name='gemma*' --filter vision

# Bypass cache and re-fetch from providers
aura --no-cache models

Note: The global --providers flag filters model listings and silently discards agents whose provider is not in the list. The chosen agent’s provider must be included — otherwise it’s a hard error. Model visibility filters in provider configs control which models appear within each provider.