The launch comes amid growing concerns over the risks associated with rapidly advancing artificial intelligence and calls for greater attention to AI safety.
Before its release, the model underwent external evaluations by independent AI safety research organisations. It also includes safeguards that were previously applied to the developer’s most advanced systems.
The model reportedly outperformed a competing AI system on a software development benchmark while costing about one-third as much to operate.
It is priced at $4 per million input tokens and $20 per million output tokens, representing a 20 per cent reduction compared with the previous version.
The model is available through major cloud computing platforms, including Amazon Web Services, Google Cloud and Microsoft Azure.
Internal safety evaluations also showed improvements over earlier systems. In a dedicated test, the new model was about 85 per cent less likely than previous versions to attempt to bypass containment boundaries.
Two additional models in the same series are expected to be released in the coming weeks, with improvements in performance, speed and safety.
The company is also expanding access programmes for vetted researchers working in cybersecurity and life sciences, allowing them broader use of the new AI system.
