OpenAI CEO Sam Altman privately demonstrated a new AI model series codenamed "Astra" to multiple U.S. senators on Capitol Hill this week, emphasizing multi-agent collaborative capabilities for long-horizon task processing. According to The Information, Altman met with U.S. Senators Raphael Warnock and Bernie Moreno on July 29, and was scheduled to meet with Senator Mark Warner, the top Democrat on the Senate Intelligence Committee. During closed-door meetings, Altman gave live demonstrations of Astra's core functionality, showing how it breaks down complex tasks and orchestrates multiple AI agents to complete them collaboratively. Tech outlet eWeek reported that he also met concurrently with Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick, previewing the next-generation model for the White House economic team. Nvidia CEO Jensen Huang also met with lawmakers—including Senator Warner—during the same period to discuss U.S. leadership in AI and model access issues. The Astra model series supports multiple AI agents working in parallel within a unified task framework, sustaining complex tasks over extended time horizons. According to eWeek, Astra assisted in disproving a core assertion of the 80-year-old Erdős unit distance conjecture during internal testing. Altman declined to disclose what new capabilities Astra possesses and did not confirm a public release timeline.