OpenAI announced a new model GPT‑6 Astra
- OpenAI has released GPT‑6 Astra
- The model surpasses GPT‑5 in accuracy
- Best results in cybersecurity
- Reduced number of hallucinations and violations
OpenAI has officially announced the launch of its latest language model – GPT‑6 Astra. This is already the sixth generation in the line of transformers designed for text generation and solving a wide range of tasks, from coding to data analytics.
Improvements compared to GPT‑5
According to the results of internal and external benchmarks, GPT‑6 Astra demonstrates a significant increase in accuracy. The model less frequently provides incorrect information, which reduces the risk of spreading misinformation. Additionally, there is a decrease in 'hallucinations' — situations where the system generates plausible but inaccurate content.
Computer work tests, including script writing, automation of routine tasks, and programming support, show that Astra outperforms its predecessor in most performance and speed criteria.
Cybersecurity and compliance
Particular attention in the new tests was given to the model's ability to recognize and respond to cybersecurity threats. GPT‑6 Astra proved to be more reliable in detecting vulnerabilities, analyzing logs, and formulating recommendations for system protection. At the same time, the model almost never violates established restrictions, which enhances its suitability for use in sensitive and regulated areas.
Developers also noted that Astra significantly less often attempts to bypass built-in restrictions, which reduces the risk of unintentionally generating dangerous or unacceptable content.
Detailed technical specifications, a list of supported features, and model limitations are available in the official product card on the OpenAI website. The card includes information about the model size, datasets used, and training methods.
The launch of GPT‑6 Astra occurs in the context of increasing competition in the field of artificial intelligence, where large tech companies are striving to offer safer and more reliable solutions. For users, this may mean more accurate answers, enhanced protection against errors, and more responsible interaction with AI.
OpenAI emphasizes that further development of the model will be accompanied by continuous monitoring of its behavior and updates aimed at minimizing risks and improving the quality of text generation.
Source: N+1
