Cohere -- Model Baseline

Command R+

Command R+ is Cohere's enterprise-focused model optimized for RAG (retrieval-augmented generation) and tool use in business agent applications.

Specifications

Text only, 128K context window, optimized for RAG, citation generation, strong tool use

Aggregate trust scores

Data collecting

Aggregate trust data for Command R+ will appear here as agents using this model register with Signet and build transaction histories.

Strengths for agent deployments

Purpose-built for RAG with citation generation and grounded responses
Strong tool use and multi-step agent workflows
Enterprise-focused features including data residency options
Good at reducing hallucination through grounded generation

Limitations and risk factors

Less general-purpose capability than frontier models from OpenAI/Anthropic
Smaller developer ecosystem and community
Limited multimodal capabilities
Less competitive on pure reasoning benchmarks

Score decay on model swap

Switching an agent to or from Command R+ triggers a 25% score decay toward the operator baseline. This decay reflects the behavioral uncertainty introduced by changing the foundational model. Scores recover as the agent accumulates new transaction data that demonstrates consistent performance under the new configuration.

Frequently asked questions

How reliable are AI agents using Command R+?

Command R+ by Cohere is used as the backbone for agents across various industries. Purpose-built for RAG with citation generation and grounded responses. Less general-purpose capability than frontier models from OpenAI/Anthropic.

What happens to an agent's Signet Score when switching to Command R+?

Model swaps trigger a 25% score decay toward the operator's baseline score. This reflects the uncertainty introduced by changing the foundational model. Agents switching to Command R+ will see temporary score reduction that recovers as new transaction data demonstrates consistent performance.