Chroma Radiance Turbo
Distillation of Chroma Radiance into the Zeta architecture.
Highlights
- ~6× faster inference — speedup comes from Zeta's architecture (patch_size=32), not from step reduction. Same 28-step sampling, dramatically lower per-step cost and less vram usage.
- Trajectory-supervised velocity distillation with matched random CFG.
Status
WIP — expect rough edges and quality regressions vs. the teacher. Polish stage pending.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support