OpenAI’s Orion, said to be 100x more powerful than GPT-4, is coming

Key Points 

• OpenAI plans to launch its next big AI model, Orion, in December, initially restricting access to select industry partners.

• Orion is said to be up to 100 times more powerful than GPT-4, marking a significant advancement in AI capabilities.

♨️News - OpenAI is preparing to launch its next big AI model, Orion, by December.

Unlike the last two models, GPT-4o and o1, Orion won’t roll out widely on ChatGPT at first. Instead, OpenAI plans to provide access primarily to close industry partners, allowing these companies to develop new products and features with the model. 

Microsoft engineers are already getting ready to host Orion on Azure as early as November, according to a source familiar with the plans.

🤓What to expect - Orion has been hinted to be up to 100 times more powerful than GPT-4, making it an impressive step forward. It is, however, separate from the o1 reasoning model released in September. And while it’s seen as the successor to GPT-4 within OpenAI, there’s no confirmation yet if Orion will be branded as GPT-5. 

OpenAI’s longer-term goal is to eventually integrate its models into something even more capable, potentially reaching what’s known as artificial general intelligence (AGI).

Meta's new Llama 3.2 models bring powerful AI to smartphones

Key Points 

• Meta has developed smaller Llama 3.2 AI models for smartphones and tablets.

• The new models operate up to four times faster and use less than half the memory of earlier versions.

• This innovation allows smartphones to process sensitive information quickly and privately, addressing growing concerns over data privacy and AI transparency.

🚀News - Meta has created smaller versions of its Llama artificial intelligence models that can run on smartphones and tablets, moving AI capabilities beyond data centers. 

The new Llama 3.2 models, specifically the 1B and 3B versions, operate up to four times faster and use less than half the memory of earlier models. Testing shows these smaller versions perform almost as well as their larger counterparts. Meta achieved this by using a technique called quantization, which simplifies the math behind AI.

Tests on OnePlus 12 Android phones revealed that these models are 56% smaller and process text more than twice as fast, handling up to 8,000 characters—suitable for most mobile apps.

🧙🏻‍♂️Why this matters - While cloud-based AI will still tackle complex tasks, these new models suggest a future where smartphones can quickly and privately handle sensitive information. 

Moreover, this development comes at a time when tech companies are under pressure regarding data privacy and AI transparency. By allowing AI tools to run directly on phones, Meta not only addresses these concerns but also presents an opportunity for developers to create applications that blend the accessibility of mobile platforms with the sophistication of advanced AI.