Models
The models we use.
SoyLic Kui is the main model. You can choose Flash, Pro or Luna in the chat, and OCR and embeddings run on our servers.
Language models
They come from third parties. They receive the text of the conversation and only the documents you uploaded or selected.
- SoyLic KuiClaude Haiku (Anthropic) · United States
- Assigned by SoyLic. It is the main model: connectors, titles, memory and tabular review, and backup for the chat.
- SoyLic FlashDeepSeek · China
- You choose it in the chat. It also reads photos, long documents and scans.
- SoyLic ProDeepSeek · China
- You choose it in the chat, for conversations that need more reasoning.
- SoyLic LunaOpenAI · United States
- You choose it in the chat.
On our servers
They run on SoyLic infrastructure. What they process does not go to a model provider.
- Local OCRSoyLic
- Reads scanned pages and extracts their text.
- Local embeddingsSoyLic
- Turn text into vectors to search laws and theses by meaning.
Google data (Drive, Gmail and Calendar) only reaches the model SoyLic assigns to the connectors: SoyLic Kui (Claude Haiku, from Anthropic), with SoyLic Luna (OpenAI) as backup. It never reaches DeepSeek: SoyLic Flash and SoyLic Pro cannot use the Google connectors.
SoyLic does not use your data to train its own models.
Each provider's policies: DeepSeek, Anthropic, OpenAI.
This page is part of the Privacy Notice (in Spanish). Updated on November 2, 2026.