OpenAI rolled out Canvas support for o1 model with React and HTML rendering

OpenAI recently shipped an update to their Canvas feature, enabling the Canvas tool for the o1 model. Previously, Canvas was only accessible with GPT-4o, but now OpenAI’s latest and smartest model has access to this tool as well. Users have already reported significantly improved results in terms of code and visuals, noting that o1 now performs on par with Claude Sonnet, which had previously led in this area.

However, the Canvas feature is not yet available to o1 Pro and is only accessible on the standard o1 model. Despite this, some users have found a workaround: they ask o1 Pro to generate content for Canvas, then switch to standard o1 to plot it on Canvas.

Additionally, OpenAI has released rendering capabilities for HTML and React code, making these features accessible to all users, including those on the free tier. These capabilities, which had been in testing for a while, allow users to visualize React applications and HTML pages and experiment with UI elements interactively.

Another notable update is the introduction of Canvas support to the ChatGPT desktop app for macOS across all subscription tiers. Users can now invoke the Canvas option by tapping the enter /slash command in the prompt area.

Discussions have also surfaced about the upcoming o3 mini model launch. Sam Altman recently asked the AI community for feedback on how many queries per day they would find sufficient, suggesting hundreds of daily queries might be available. Following community backlash, he confirmed that users would indeed be able to query o3 mini hundreds of times per day.

Finally, there were hints that the previously released Operator was just the first of OpenAI’s agent-based tools, with more agents expected to launch in the future. Speculation suggests this could tie into the anticipated software engineering agent, as hinted by the ChatGPT Twitter account last year with references to “software making software.” Exciting developments lie ahead!