Reasoning control
Choose the depth of deliberation a task deserves.
Models
Open models built to carry work.
GRaPE and CRePE are SLAI's model families for controllable reasoning, deliberate action, tools, coding, and project exploration.
General Reasoning Agent for Project Exploration
GRaPE
GRaPE is the general-purpose model family. It is trained around the work SLAI products ask models to do: understand mixed inputs, reason through plans, use tools, write artifacts, and stay useful over longer tasks.
View on Hugging Face
Current release
GRaPE 2
GRaPE 2 brought SLAI's general reasoning family up to the competition, with stronger project exploration, tool use, coding, multimodal work, and controllable effort.
Code Reasoning Expert for Project Exploration
CRePE
CRePE focuses the same project-exploration premise on software work. It is designed for agents that must inspect a codebase, reason about changes, use tools, and continue until the result is complete.
Follow SLAI models
Shared direction
Control is a capability.
Open access matters more when people can also shape how a model reasons and works.
Work effort
Control how far an agent carries the work, separately from how long it thinks.
Tool use
Build models for active project exploration instead of isolated responses.
Multimodal work
Every GRaPE model is built for vision, treating images and text as parts of the same task.
Agentic coding
Read, reason, edit, test, and continue across real software projects.