encode_bench

Can a model generate base64 that decodes into the requested artifact? (proxy for capability)

loading…
Loading battery and per-model results…