Kleros floats benchmarking GPT-6 Astra as a V2 juror agent

Kleros proposed putting OpenAI's GPT-6 Astra to the test on its platform: read all the evidence in a real Kleros case and vote with the majority of jurors. The tweet, from Kleros's official account, invites anyone to plug GPT-6 Astra into a juror agent on Kleros V2 and run the benchmark today.

The pitch positions the scenario as a useful benchmark for the model's intelligence and alignment claims, tying it directly to Kleros's push toward AI agent jurors that autonomously read case evidence and vote.