☀️ 🌙

Inference-Time Scaling

Explore how ITS algorithms enhance model performance through computational scaling at inference time

📚

Guided Demo

Explore pre-loaded scenarios showing representative ITS improvements across different algorithms and providers.

  • Pre-loaded scenarios with representative results
  • Compare baseline vs ITS side-by-side
  • View algorithm traces and metrics
  • No backend required - instant loading
  • Perfect for presentations and demos
Explore Scenarios →

Interactive Demo

Run live comparisons with your own questions and configurations against hosted models.

  • Ask your own questions
  • Choose models and algorithms
  • Adjust compute budgets
  • See ITS improvements in real time
  • Compare use cases (improve, match frontier)
Start Experimenting →

How to Use in Red Hat AI (Work in Progress)

ITS is officially available in Red Hat OpenShift AI through the Red Hat AI Python Index — built and delivered by Red Hat, optimized for hardware accelerators, and tightly coupled to Red Hat AI Base Images for a secure, reliable AI supply chain.

📦

Red Hat AI Python Index

Install the its_hub SDK directly in your RHOAI workbench. Base images are pre-configured to use the index — packages are built by Red Hat, tested for platform integrity, and optimized for CUDA, ROCm, and CPU.

pip install its_hub
🖥️

RHOAI Model Serving

Deploy models with the RHOAI vLLM serving runtime (Granite, Llama, etc.) and point its_hub at the serving endpoint. Fully on-cluster, no external API calls needed.

endpoint = "https://<model-route>/v1"
🔮

Coming Soon: AI Gateway

ITS will be integrated into the Red Hat AI Gateway, providing inference-time scaling as a managed service — no SDK installation required.

Red Hat AI Gateway — stay tuned