AI Models
Which models power Valarchi's planning agents, and why.
Supported models
Valarchi currently uses Claude models from Anthropic across all planning agents — PRD, App Structure, Wireframes, Tech Plan, and Codebase Analysis.
Anthropic — Claude Sonnet 4.6
Context window
1,000,000 tokens
Provider
Anthropic
All agent invocations use Claude Sonnet 4.6 via the Anthropic API. During your trial, Valarchi provides the API access. When you bring your own key, you supply a key from console.anthropic.com.
Why Claude?
We evaluated multiple models internally before building Valarchi. Claude Sonnet 4.6 stood out on two dimensions that matter most for planning agents: latency and output quality. In our internal testing, it produced consistently better planning artifacts — faster, more detailed, and grounded in the actual product context — compared to other models we tested.
The 1M token context window is also directly useful here. A single planning session can involve a codebase snapshot, a full PRD, app structure, and wireframe references all at once. Having the entire milestone context in one call — without chunking or summarising — means the agents produce more consistent output across steps.
Other model providers
If support for other model providers is important to you, contact us and let us know — it helps us understand demand.