OpenAI’s unreleased Astra can coordinate 16 agents at once, The New Stack reports
The New Stack says Astra can run full experiments and may have hit OpenAI’s highest cybersecurity risk threshold.
TLDR
The New Stack describes OpenAI’s unreleased Astra as capable of doing a researcher’s week of work. It reports that the model can run full experiments and coordinate 16 agents at once, and says Astra may have hit OpenAI’s highest cybersecurity risk threshold.
Combined views
486
1 Source, first seen ago
1 likes