OpenAI Shakes Up the AI World with GPT-5.6 and a Wave of Departures

OpenAI and Meta: Major Advances in Language Models

Recently, the world of artificial intelligence has been shaken by the launch of two significant language models: GPT-5.6 from OpenAI and Muse Spark 1.1 from Meta. Although the launch day is not ideal for fully assessing these innovations, the announcements surrounding them have generated considerable interest.

OpenAI Unveils Its Innovations

OpenAI made waves with the launch of a particularly advanced voice model, accompanied by a striking advertising campaign. According to M.G. Siegler, this model could mark a major turning point in computing, where voice could become a primary method of interaction. This advancement is promising for OpenAI's hardware ambitions, suggesting that the new capabilities of GPT-Live could pave the way for new technological devices.

The launch of GPT-Live was just part of a series of ambitious announcements. OpenAI integrated Codex into its desktop application, introduced the ChatGPT Work agent, and retired the Atlas browser. The highlight was the launch of GPT-5.6, available in three versions: Luna, Terra, and Sol, each offering increasing performance.

In a blog post, OpenAI emphasized that Sol has outperformed Anthropic's Claude Fable in several benchmarks, including the Agents' Last Exam and the Artificial Analysis Coding Agent Index. Early user feedback has been enthusiastic. Katie Parrott from Every described GPT-5.6 as her favorite model for collaboration, although she still trusts Fable for certain uses.

She added that this model represents a significant shift in the capabilities of models for everyday knowledge work, being fast and clever enough to adapt to users' changing needs.

Box CEO Aaron Levie also expressed his excitement, stating that Sol represents a major advancement over GPT-5.5, particularly for complex tasks requiring in-depth analysis.

Sol versus Fable: Mixed Opinions

Opinions vary regarding the comparison between Sol and Fable. Mitchell Hashimoto, founder of HashiCorp, expressed a preference for Sol, finding it faster and more efficient for planning and judgment. However, he continues to use Fable for specific tasks requiring precise debugging.

Dylan Field, CEO of Figma, advised against direct comparisons, deeming them somewhat irrelevant. Nevertheless, he acknowledged the quality of Sol.

For Ethan Mollick, a professor at the Wharton School, both Sol and Fable represent significant advancements over previous models, creating a notable gap with other AIs. He considers these models essential for work requiring heightened intelligence.

Turbulence at OpenAI

Despite the excitement surrounding GPT-5.6, OpenAI is facing internal turbulence. Fidji Simo, the company's number two, announced her departure, citing a worsening of her chronic illness, postural orthostatic tachycardia syndrome. She will continue to advise the company.

Simo joins a growing list of executives who have left OpenAI this year, including Joshua Achiam, Jerry Tworek, Kevin Weil, Barret Zoph, Joanne Jang, and Max Schwarzer. These departures raise questions about the company's stability as it continues to advance in the field of AI.