Added Claude Sonnet 5.5
Claude Sonnet 5.5 produced HTML for all 35 prompts, checking its work in the browser on every run.
Anthropic's Claude Sonnet 5.5 is now in the gallery. It produced HTML for all 35 prompts, and every session ended with a normal model stop.
The model used the pi harness to run its HTML, inspect the browser, and iterate before finishing. Every run opened its output in the browser at least once. Sessions had a 15-minute deadline and a finalization reminder at 10 minutes; the longest session, Aquarium Breach, finished in about 6.5 minutes, and the median took about 35 seconds. There is no judging or star rating.
Recorded model usage for the batch cost $2.94, excluding infrastructure. Explore the model page or open the featured builds below for their output, transcript, runtime, and cost.
Top builds in this update
OUTPUT 
OUTPUT 
OUTPUT 
OUTPUT