OpenAI Unveils GPT-6.1 Sol, a More Cost-Efficient AI Model

OpenAI, a leading US artificial intelligence (AI) research company, has unveiled a new AI model, GPT-6.1 Sol, at its annual technology conference, Dev Day.

Launched just a week after GPT-6 Sol, the new model delivers performance close to that of the company’s flagship GPT-6 Astra in coding, computer operation and the automation of complex enterprise tasks.

The most significant technological feature of the new version is its ability to deliver intelligence nearly on par with Astra while reducing input and output token costs to just one-fifth of those of the standard model.

Security Concerns and Performance Improvements

OpenAI has temporarily cancelled the release of its highly anticipated GPT-6.1 Astra after internal testing revealed complex security issues. Researchers recommended the decision after tests identified a tendency for the model to mislead users and a risk that it might independently perform tasks without authorization.

However, GPT-6.1 Sol brings substantial improvements in factual accuracy and complex problem-solving. Its rate of general factual errors has been reduced from 11.4 per cent to 7.7 per cent, leaving it just 1.9 percentage points behind the flagship Astra model in performance.

Security Features and Availability

GPT-6.1 Sol is also considerably more reliable than its predecessors in acknowledging its technical limitations and adhering to strict security requirements. It can carefully execute multi-step workflows in accordance with users’ original instructions without taking unauthorized actions.

The new AI model is already available to Plus, Pro, Business, Enterprise and Edu customers through the ChatGPT Work and Codex platforms. However, it will take some more time before it becomes available in the standard chat mode.

//DBTech/BMT/OR//