“Our domain-specific Asta v0 (which uses a mixture of LLMs) led at 53.0%, ~10 points above ReAct-gpt-5, though at higher engineering and runtime costs. … Note: Asta v0 is an experimental agent not used in the production version of Asta today.”
Sounds like more BS AI hype from techno grifters.