Models for the masses

Letter from the editor
Sabrina Ortiz, Senior Editor, AI

September is typically referred to as Techtember around here because of the deluge of hardware launches. But this week, AI news kept us just as busy with the release of some of the most anticipated LLMs from OpenAI and Anthropic, revealing a bigger trend in where the industry is going. But first, let’s take a speed run at the week’s biggest announcements (buckle up).

OpenAI set the tone for the week with the release of two new open-weight models. This was major news because it was OpenAI's first public release of model weights since GPT-2 in 2019, and likely in response to the public’s requests. The biggest beneficiaries are developers who can now build on, fine-tune, and even run the models locally for added security.

The same day, Anthropic announced its highly anticipated Claude Opus 4.1 model. When Claude Opus 4 launched in May, Anthropic called it the "best coding model in the world," though our testing found it didn’t beat its free sibling model, Claude 4 Sonnet. Still, Opus 4.1 performed higher on SWE-bench Verified, which evaluates LLMs' abilities to solve real-world software engineering tasks sourced from GitHub. People have already built some interactive coding projects using Opus 4.1 and a conversational prompt in minutes -- a task that could take hours.

Google also got in on the action. Beyond unveiling its Genie 3 world model on Wednesday, the company offered its Google AI Pro plan for free to students 18 or older in the US, Japan, Indonesia, Korea, or Brazil. Normally $20 a month, the plan bundles all of Google's AI offerings, including the best models in Gemini and standalone tools such as NotebookLM, Jules, and Gemini 2.5 Pro.

Perhaps the biggest launch of the week was OpenAI’s GPT–5, a next-generation family of models the company touts as its fastest, smartest, and most capable yet. Already available in Copilot, GPT-5’s biggest appeal is that it combines a smart model for most queries with a deeper reasoning model for harder problems, choosing which to use based on the context of the prompt. In a huge break from OpenAI’s norm, even free-tier users can access it.

And it certainly appears capable. I watched a live demo of GPT-5 in which a user created a fully functional web app with interactive elements from just a text prompt. As someone who recently learned how to build webpages, that would have taken me hours to stylize using JavaScript and CSS.

Also notable: Anthropic made it easier to trigger a security review in Code Claude, OpenAI rolled out non-GPT-5 updates to ChatGPT, and Google embedded AI agents deeper into its data stack. 

These launches made one thing clear: As AI adoption falters and hype cycle promises wane, companies are making more advanced models accessible to more people, hoping for a sea change. This looks like a potential new wave of AI democratization; instead of gatekeeping the most advanced models for the highest-paying customer, major players are letting everyone try them, whether that’s a developer looking for open-weight models to build on or a free user looking for the best answer possible.

Giving developers and the general public expanded access to the latest models and tools should unleash a future wave of innovation and make it easier than ever to build solutions. Only time will tell, but we can all agree that it is an exciting time for AI. 

Follow me on X, Instagram, and LinkedIn to see my latest coverage.