September has been a quieter month in terms of AI development but that likely means there's something brewing; here are some of the biggest stories this month.

AI passes CFA III

Earning the coveted Chartered Financial Analyst (CFA) test usually requires people to spend about 1,000 study hours spread across several years to clear its rigorous three-level exam. However, recent research shows that the technology driving many AI models has become capable of passing even the toughest Level III practice tests in just minutes. Previous models were able to pass Level I and II with ease but struggled with the final level due to the essay questions.

Researchers at the New York University Stern School of Business conducted the study and proved frontier models such as o4-mini, Gemini 2.5 Pro, and Claude Opus were able to pass the test with ease.

Researchers discover exotic state of matter

Using a 58-qubit superconducting quantum processor, researchers from the Technical University of Munich (TUM), Princeton University, and Google Quantum AI successfully created a Floquet topologically ordered state – a phase that had been theorised but never experimentally observed. They captured direct images of the distinct edge-directed movements and introduced a new algorithm to explore the system’s topological features. This enabled them to observe the dynamic "transmutation" of exotic particles, a defining phenomenon long predicted for such quantum states. All a bit complicated but very impressive!

The article by TUM stated: "This work opens the door to a new era of quantum simulation, where quantum computers become laboratories for studying the vast and largely unexplored landscape of out-of-equilibrium quantum matter. The insights gained from these studies could have far-reaching implications, from understanding fundamental physics to designing next-generation quantum technologies." Quantum technology combined with artificial intelligence will no doubt have a large impact on scientific discoveries in the future.

Anthropic: coding leaders

US startup Anthropic on Monday announced the launch of its new generative artificial intelligence model, Claude Sonnet 4.5, which it says is the world's best for computer programming. While trailing OpenAI in terms of users and name recognition, Anthropic has been considered the top performer for several months in generative AI for computer coding. Claude Sonnet 4.5 achieved the highest score when tested by the independent evaluation system SWE-Bench Verified, developed by researchers from Princeton and Stanford universities.

OpenAI's Pulse

September also saw OpenAI roll out ChatGPT Pulse, a new preview feature for Pro users that shifts the language model functions from reactive to proactive. It runs background research overnight, pulling from past chats, memory, and connected apps to deliver morning cards with tailored updates, event reminders, and goal tracking. The move signals OpenAI’s push to make ChatGPT feel less like a tool you prompt and more like an assistant that anticipates your needs.

New AI powered Ray-Bans

Meta has launched the Ray-Ban Display, its most advanced smart eyewear to date. Priced at $799, the glasses feature a discreet in-lens display for real-time captions, translations, and notifications, powered by Meta AI. Control is achieved through the accompanying Meta Neural Band, an EMG (electromyography) wristband that translates subtle muscle movements into commands. This combination aims to keep users connected without distraction. The Ray-Ban Display glasses are now available in select US retail locations, with demos offered at Meta’s pop-up shops in New York and Las Vegas, and will likely be in Europe soon enough.

World's first video reasoning model

This one will likely have a huge impact in the world of cinema; Luma AI has unveiled Ray3, its next-generation video model that "thinks" in visuals. With 16-bit HDR video generation, Ray3 can interpret prompts, evaluate its own output, and refine internally – turning ideas into cinematic clips. Its new Draft Mode speeds up concept development by letting creators experiment with multiple visuals at once, then refine them into high-quality final cuts. Watch the video below and you will see how realistic it all looks!

Despite it being a slower month, the AI sphere is moving at an incredible rate and the world is changing, therefore it is critical to stay up to date with all developments in the field. You can read the previous months' AI reviews here.