Frontier performance,
without the frontier price
Frontier performance at a fundamentally lower cost
Hal 1.0
Discover Hal 1.0Capability
Designed for complex
reasoning, agent workflows,
and software tasks.
Significant savings with prompt caching (up to 90%) on repeated prompts.
Hal is best suited for:
Multi-step agents and tool calling
Deep code generation and review
Complex reasoning across long documents
Comparable to Opus 5.5
and GPT-6 Astra
Clarke 1.0
Discover Clarke 1.0Performance
Strong performance with the best balance of speed and cost.
Significant savings with prompt caching (up to 90%) on repeated prompts.
Clarke is best suited for:
RAG and knowledge assistants
Copilots, support chat and tool calls
Code assistance and structured outputs
Comparable to Sonnet 5.5
and GPT-6 Sol
Tycho 1.0
Discover Tycho 1.0Scale
Ultra-efficient inference for high-throughput applications, early development, and lightweight agents.
Significant savings with prompt caching (up to 90%) on repeated prompts.
Tycho is best suited for:
Classification and intent detection
Extraction and summarization
Routing and support ticket triage
Comparable to Anthropic Haiku 5.5
and GPT-6 Luna
A per-token difference doesn't stay small for long.
Start running inference
on Radium
Resources for teams evaluating, integrating, and operating Radium
One line of code to switch.
A different class of performance.
Swap OpenAI for Radium in your API call. Get $25 in credit. No card required.