A place for knowledge.
Make reference information part of the agent your team carries.
A world of knowledge is only useful
when it’s within reach.
Atlas is one application of the Zaro approach: a reference and retrieval agent designed for local devices.
Atlas pairs your knowledge with a compact model. Its default model is a curated Qwen selection, not a Zaro-trained model. Need task-specific model adaptation? We can scope that as a separate project.
Make reference information part of the agent your team carries.
Bring retrieval and local language intelligence into one purposeful experience.
Scope a configuration for your devices and the places your people work.
Our default choice for Atlas, based on Qwen2.5-1.5B-Instruct. Multilingual instruction-following, a 32k-token context window, and a 986 MB download.
Model download
Context window
Download size is not the model’s total memory requirement. We’ll confirm compatibility with your devices during deployment planning.