Why Holarch Uses Claude

Holarch's AI features currently use Claude (Anthropic): in Holarch's own tests, only the Claude models met its quality bar.

• The test: real requests from Holarch projects: requirement reviews with seeded defects and duplicates, rewrites, decompositions, flow diagrams checked by Holarch's parser, trade criteria, risk candidates, and Ask AI questions with known answers, including “not yet” questions. Each request is sent three times.
• The bar: at least 90% of tasks fully usable, every “not yet” question answered truthfully, every seeded defect and duplicate caught, and every answer valid for Holarch to read.
• Result (October 2026): Claude Haiku 5.5, Sonnet 5.5 and Opus 5.5 met the bar. The OpenAI and Google models tested did not yet (Google's API also did not accept keys at the time).
• What changes it: new models are tested as they appear; a provider is offered again when its models meet the bar. A key you added earlier for another provider is kept and not used.
• Whatever the model, every AI change is a preview you review and apply as one undo step.
• To ask for another model, use Send Feedback in the account menu.

Related

AI Settings and Keys · Which AI Key Is Used