Native video inference beats sampled frames, but OpenRouter providers don't support it yet

spillai · x · 2026-09-08

Responding to the egocentric VLM eval thread, spillai (vlmrun) notes that evaluating via sampled images instead of native video likely hurts performance, and that OpenRouter providers generally don't support native video out of the box — he saw real discrepancies while running evals. His vlmrun gateway will host more VLMs/world models for robotics evals, exposing fps, dynamic sampling, and resolution controls relevant to these use cases.

Original post →

More from Models

Models channel →