Skip to benchmark content
MotionBenchHosted by baz.studio
July diagnostic pilot33 routes · diagnostic datasetJuly diagnostic pilot
openai · O+C

GPT 5.4

gpt-5.4
Diagnostic tierC
Full video
Human assessment
Valid but small and visually conservative in sampled frames.

Historical diagnostic evidence; not a durable capability claim.

GPT 5.4 — MotionBench