OpenAI launches Astra, its powerful (and controversial) new model
By Pradhyuman,
OpenAI released its new model, Astra, on Thursday. The company made Astra available first to users of its Daybreak cybersecurity program, with plans to roll it out to paid accounts and its API over the next week. OpenAI president Greg Brockman stated that Astra is the company's most intelligent model.
OpenAI stated that Astra helps defenders patch weaknesses by identifying zero-day exploits. Benchmark tests provided by OpenAI show Astra scored higher at finding bugs than OpenAI's Sol and Anthropic's Fable.
Astra uses a reasoning technique called opaque recurrence, which obscures chain-of-thought monitoring. OpenAI chief scientist Jakub Pachocki stated that monitoring model reasoning becomes more challenging as capabilities increase. When asked about artificial general intelligence, Brockman stated that contractual AGI triggers are no longer relevant following contract changes with Microsoft.
Maintained by Pradhyuman.
Filed under: OpenAI, Models & Research, Products & Agents