Saro 1B
In developmentYour knowledge.
More to reason with.
Saro / Research programme
Scale
~1B parameters
Focus
Document knowledge
Status
In development
Built around
a specific question.
Can a small model work reliably with knowledge that was never part of its original training?
Saro 1B is our model programme for reasoning with user-chosen knowledge: the documents a person gives it, kept available as it works. Our goal is to help it connect ideas, draw conclusions and solve problems using that knowledge.
Our goal is useful capability at a practical size. Accuracy, appropriate abstention, memory use and inference speed all matter.
Where we are.
The model is in development. Early versions already hold conversations in English and Spanish and answer questions about documents they have been given. We are improving accuracy before a public release. Model weights and a technical report are not yet available.
Can I use Saro 1B now?
Not yet. This page introduces the research programme. We will publish availability and validated capabilities when the model is ready.
How can I collaborate?
For research collaboration and conversations about document-intensive use cases, write to hello@ideofox.com.