Local Model Benchmarks

Local test machine: Apple M5 Pro · 64 GB unified memoryRun window 29 September — 18 October 202622 recorded runs

Challenge prompts

Read and copy the Digger 3D, Hardball 3D, Obby 3D and Wild West Showdown 3D prompts and the three task-loop briefs, or open the original Whiteout Protocol prompt. Then compare what each setup produced.

Available to read and copy

Original source

Not published

Agent transcripts and private runner prompts are outside the published subset. Sharing a challenge prompt does not remove differences in scaffolds, agents or tools.