ChatGPT Pro's Pricey Leap

OpenAI has launched ChatGPT Pro, a $200-per-month subscription designed for power users who want to push AI to its full potential. The centerpiece is o1 Pro Mode, an enhanced version of the o1 reasoning model, which uses extra compute power to tackle complex tasks like coding, math, and legal analysis with more precision.

The o1 model itself is a leap forward, reasoning step by step through tasks and now supporting image uploads—a feature absent from earlier versions. OpenAI claims the Pro Mode slashes errors on difficult questions by 75%, though oddly, some benchmarks still favor the preview model. It’s a trade-off between more advanced reasoning and slightly slower response times.

Priced at ten times the cost of ChatGPT Plus, the Pro plan is drawing mixed reactions. To sweeten the deal, OpenAI is offering unlimited access to GPT-4o and Advanced Voice Mode, along with grants for free subscriptions to select researchers. For OpenAI, which expects to lose $5 billion this year, it’s a bold step to drive revenue while offering something unique for AI enthusiasts.

For most users, though, the question lingers: is $200 a month worth smarter, slower AI? OpenAI is betting that power users will say yes.

Key Features of ChatGPT Pro:

• Access to o1 Pro Mode for advanced reasoning and fewer errors.

• Unlimited GPT-4o access and Advanced Voice Mode.

• Support for image uploads and complex problem-solving.

• Grants available for researchers in select fields.

Aurora Takes the Stage—Then Disappears

X, formerly known as Twitter, launched a new AI image generator called Aurora on Saturday, but its debut was as fleeting as a tweet. After a few hours of going live on the Grok assistant, Aurora mysteriously vanished for many users, leaving questions hanging in the air.

Aurora is already making waves with its ability to produce stunning, photorealistic images. From lifelike landscapes to detailed still lifes, the results are impressively polished for a beta-stage tool. While it struggles with common AI pitfalls like hands, its visual output is a big step forward, hinting at even greater potential with further refinement.

Unlike most image generators, Aurora has few restrictions. Need a picture of Mickey Mouse? No problem. A controversial political figure? Sure thing. But while Aurora didn’t allow nudity in early tests, it wasn’t shy about generating graphic content, sparking mixed reactions online.

Despite its photorealistic chops, Aurora has its quirks—think objects melting together and the classic AI struggle with hands. Still, its landscapes and still lifes shine, suggesting plenty of promise once the bugs are ironed out.

Aurora’s release follows Grok’s shift to a free model, allowing users to generate up to three images daily. Meanwhile, xAI closed a $6 billion funding round and teased big plans, including a standalone app for Grok and the potential release of Grok 3. Looks like Aurora might be just the beginning.

Amazon's AI Ambitions: Chipping Away at the Competition

Amazon is making a big splash in the AI world, and investors are loving it. This week, the company unveiled its Trainium2 chips, designed to rival Nvidia and AMD for AI-heavy computing. But that’s not all—Amazon also introduced the Rainier supercomputer, built to take on Elon Musk’s Cortex and Colossus systems.

The company also launched six foundational AI models under the Nova umbrella. These large language models (LLMs) promise big savings—up to 75% cheaper than competitors—and support more than 200 languages, with some even handling multimodal outputs like images and video. Experts are calling this a bold move.

The Nova models shift Amazon from being a mere broker of AI tools to a full-on content creator. Meanwhile, Trainium2 chips, boasting 30–40% better price performance than GPUs, have already drawn commitments from Apple and Anthropic. By combining hardware and software innovation, Amazon is positioning itself as an end-to-end AI powerhouse.

Amazon says this is only the beginning, with plans to roll out even more advanced models and reduce costs further in 2025. With these moves, Amazon isn’t just hosting AI; it’s aiming to own it, setting the stage for what could be a massive industry shake-up.

Meta Debuts Llama 3.3

Meta has unveiled Llama 3.3 70B, a new AI model offering the power of its predecessor, Llama 3.1 405B, at a lower cost. Ahmad Al-Dahle, Meta’s VP of generative AI, says the improvement comes from advanced post-training techniques. Benchmarks show it outperforming competitors like Google’s Gemini 1.5 Pro and OpenAI’s GPT-4o.

The model is available on platforms like Hugging Face, though large platforms with over 700 million users still need a special license. Despite restrictions, Llama’s semi-open nature has made it a hit, with over 650 million downloads so far.

Meta is using Llama internally too. Its Meta AI assistant, powered entirely by Llama models, now has nearly 600 million users monthly. CEO Mark Zuckerberg believes it’s on track to become the world’s most-used AI assistant.

Meta faces regulatory hurdles in the EU, particularly around GDPR and AI training practices. Still, the company is pushing forward, announcing a $10 billion AI data center in Louisiana to train Llama 4, which will need 10x more compute power than Llama 3.

With AI development driving up capital spending by 33% to $8.5 billion in Q2 2024, Meta’s bets on AI are costly. But with Llama 3.3 impressing and Llama 4 in the works, the company is determined to dominate the AI landscape.