Following the launch of GPT-6 Astra and the update to version 6.1, OpenAI has introduced GPT-6.2 Astra and Sol. Two new models that certainly catch the eye, since the idea is that we’ll be able to delegate more complex work and do so using fewer resources. A particularly interesting improvement for those of us already using ChatGPT Work or Codex in our day-to-day work.
An AI that better understands how we want to work
With Astra, OpenAI says we can give ChatGPT a screenshot of a slide and ask it to create a full presentation following its design. The system reproduces the layout, the colors and even the font size, as we can see in the launch video. The new model also performs better at spreadsheet, document, programming and web design tasks, with results that, according to the company, need considerably fewer tweaks.
Sol, for its part, is designed to handle more routine tasks. As an evolution of GPT-6 Sol, this version is substantially more efficient. In a programming test OpenAI has published, it comes close to Astra’s performance, but at a fifth of the cost.
More capability within limits that will change soon
Today’s announcement represents another step in ChatGPT’s evolution toward an AI that acts on our behalf. Both models complete increasingly demanding tasks with fewer tokens, which is certainly welcome now that ChatGPT’s limits will be combined with those of Work and Codex.
Pro subscribers can enable the Ultrafast option, which generates tokens up to 4.5 times faster, though it uses more of the quota. Both new models have been available since yesterday, November 17, in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu plans, as well as through the API.