Clarke 1.0
For best balance of capability, speed, and cost.
Balanced Performance
RAG and knowledge assistants
Answer over internal docs, policies, customer dat a, and knowledge bases
Enterprise copilots
Assist sales, support, finance, legal, and engineering teams with contextual intelligence
Agentic workflows with tool use
Call APIs, return structured outputs, execute multi-step business processes
Significant savings with prompt caching (up to 90%) on repeated prompts.
Comparable to Sonnet 4.6and GPT-5.4
Primary use cases
Answer over internal docs, policies, tickets, customer data, and knowledge bases.
Assist sales, support, finance, legal, engineering, and operations teams.
Use tools, call APIs, and execute multi-step business processes.
Generate, debug, refactor, and explain code across production teams.
Return JSON, classifications, summaries, and workflow-ready responses.
Power AI features inside SaaS, fintech, healthcare, legal, and enterprise apps.
Analyze contracts, emails, transcripts, filings, manuals, and research.
Key capabilities
The model for most production workloads
Clarke is the model to use when Tycho is too lightweight and Hal is more than the task requires. Built for everyday enterprise intelligence: capable enough for complex workflows, efficient enough for production scale.
Designed for Claude and OpenAI migration
Clarke gives teams a practical alternative to incumbent flagship endpoints. Developers keep a familiar API pattern. Product teams keep the same user experience. Finance gets a materially better cost structure.
Built for agents and tools
Clarke is designed for modern AI applications that do more than chat. Use it to call tools, return structured outputs, reason over retrieved context, and execute multi-step workflows.
Enterprise controls without model ops
Clarke gives enterprises the outcome they want from open models without requiring them to select, host, tune, monitor, and maintain the model stack themselves. Radium abstracts the complexity behind a production API.
For teams using Sonnet or GPT flagship/default-class models.
Clarke is the Radium model for teams that would otherwise default to Claude Sonnet or a flagship OpenAI endpoint. It is designed to be the first model enterprise teams test when they want to move meaningful production volume away from incumbent API providers without compromising the application experience.
Move your default AI workloads to Clarke.
Clarke is the best starting point for enterprise teams replacing Claude Sonnet or OpenAI flagship usage with Radium.