Pilot datasetOne prompt · one trial · 33 routes · not the official Core rankingRead the limits

openai · C

GPT-5.1

gpt-5.1-2025-11-13

Pilot visual tierB

Human assessment

Good fragment field and clean final hold.

One diagnostic trial. This is not a durable model capability claim.