The experiments

Every ranking traces back to a real experiment.

We run frontier models on real, multi-step work on the Recursiv platform. Each study below is the receipts: the setup, the numbers with confidence intervals, and the actual agent transcripts.

Experiment 001·Jun 4, 2026

The real cost of finishing the job

We gave the 10 top AI models the same real coding, data, reasoning, and SQL tasks, then measured what each one finished, how good it was, and what it cost.

101×

priciest, and last place

read →

Run it yourself

Every number here came from running real agentic work on Recursiv. Point the same swarm at your own tasks.

Talk to us