Supported Models

Sylva supports models from three major AI providers. Each model is identified using the provider/model-name format.

Model Catalog

Model	Tier	Use Case
`gemini/gemini-3.1-pro`	Primary (default)	High-quality analysis with large context windows
`gemini/gemini-3.1-flash`	Mini	Fast section extraction, lower cost
`gemini/gemini-3-deep-think`	Primary	Extended reasoning for complex architectures

Model	Tier	Use Case
`anthropic/claude-opus-4.6`	Primary	Premium analysis, highest accuracy
`anthropic/claude-sonnet-4.6`	Primary (default)	Best balance of speed and quality
`anthropic/claude-sonnet-5`	Primary	Latest generation, improved reasoning
`anthropic/claude-haiku-4.5`	Mini	Fast, low-cost section extraction
`anthropic/claude-haiku-3-20250519`	Mini	Legacy mini model

Model	Tier	Use Case
`openai/gpt-5.3`	Primary	Latest OpenAI flagship
`openai/gpt-5.3-codex`	Primary	Code-optimized variant
`openai/gpt-5.2`	Primary	Strong reasoning, reliable for monorepos
`openai/gpt-4o`	Primary (default)	Battle-tested, excellent accuracy-to-cost ratio
`openai/gpt-4o-mini`	Mini	Fast, cheap section extraction

When you select a model with -m, Sylva automatically assigns:

Primary model → Used for RLM analysis and conventions compilation (accuracy-critical)
Mini model → Used for section extraction (speed-critical, cost-efficient)

The mini model is always the default mini for that provider. You cannot currently override the mini model independently.

# See all available models
npx @thelogicatelier/sylva --list-models

Each provider requires its own API key. See Configuration for setup instructions.