Everything on this site.

Home

radianvector.com/

• Two paths
• What RadianVector measures
• AI system tuning parameters
• Contact

Frontier Models

radianvector.com/frontier-models/

• How hosted models are timed
• Behavior analysis
• Configuration comparisons and reports
• What you can compare

Work with us

Tailored AI analysis for growing organizations

• Teams changing their workflows
• Teams running their own models

For Venture Capital Firms

Workload evaluations for portfolio companies

• What you receive
• Engineering & tools
• See the tools at work

Engineering & tools

Engineering evidence and RadianVector capabilities

• Compare performance and behavior
• Published correctness investigation
• RadianVector capabilities
• Actual inference measurements
• See the tools at work

Open Models

radianvector.com/open-models/

• Why configuration choice matters
• Staged configuration experiments
• The deployment frontier
• The configuration decision
• Release revalidation
• Independent evidence
• Benchmarks
• Open source

Benchmarks

radianvector.com/open-models/#benchmarks

• Beyond tokens per second
• Thinking on, thinking off

Research & open work

• Research and public work
• vLLM MTP bug investigation
• inferlint
• Tutorials
• Inside an LLM
• How an LLM generates one token
• Making the sentence stand up

RadianVector

• [email protected]