Repository navigation
Current OpenAI and Claude models, with the same setup workflow. Choose GPT-6 Luna, GPT-6.1 Terra or Sol, or Claude Haiku, Sonnet or Opus 5.5. Astra and Fable remain available, and Google choices are unchanged. Superseded models no longer appear as new selections; existing configurations remain identifiable and supported while their provider serves them.
Model settings and spending follow the selected model. Thinking uses the lowest supported setting by default. Preset pricing accounts for applicable prompt-length tiers and scheduled rate changes; manually entered rates remain yours.
Validated locally: 1,364 backend tests passed (2 skipped), 584 frontend tests passed, and both builds passed. Live OpenAI/Anthropic classification and translation, Spanish DOCX, 40-row bulk retry, Arabic/Chinese PDFs, and cache reuse passed. Google live testing was excluded because its model selections were unchanged.
Deployment verification is recorded separately in docs/deployment/v1.8.2-validation.md.