The Cost of Artificial Super Intelligence is Now $200

OpenAI’s ChatGPT Pro is sparking debates on its true value. As AI becomes more integrated into our lives, the idea of ‘intelligence too cheap to meter’ feels closer than ever. But for now, as one former OpenAI exec, now in Google, Logan Kilpatrick, said: “consumer willingness to pay for AI is going to go up (a lot).” 

The next few years will reveal whether this price point is a disruption or just the start of a bigger revolution. For instance, the o1 Pro model, bundled with the $200 ChatGPT Pro subscription, has raised questions about whether the cost is truly justified. 

However, the bitter reality is that—o1 is not for everyone. 

“A small percentage of users want to use ChatGPT a TON and hit rate limits and want to pay more for more intelligence on really hard problems. The $200/month tier is good for them!” said OpenAI chief Sam Altman, clarifying that most users will be best served by the free tier or the $20/month tier. 

OpenAI believes that its o1 Pro mode delivers more reliable and comprehensive responses, particularly in fields like data science, programming, and case law analysis. It also outperforms both o1 and o1-preview on challenging ML benchmarks across math, science, and coding.

Unlocks Scientific Breakthroughs  

The model is expected to play a key role in sectors like health sciences, assisting in the search for cures for rare diseases. For instance, Derya Unutmaz, a professor at The Jackson Laboratory, revealed in a post on X that he is using o1 pro for a cancer therapy project.

Unutmaz said that o1 Pro helped him propose a groundbreaking idea to simulate T-cell exhaustion in a way that takes into account the time-based progression of the process. He said the model, instead of just looking at a single moment, proposed introducing different stress factors to the tumour at different points in time, giving an analogy of the Battle Royale game, which challenges players to adapt to escalating obstacles. 

“To be clear, o1 Pro is a totally different beast, requiring different prompting approaches and being useful in different ways from GPT-4 or Sonnet,” Matt Shumer, CEO of HyperWrite AI, said on X.

He explained that he first tests whether Sonnet can handle complex coding tasks. If Sonnet struggles or fails, he switches to o1 Pro. He described o1 Pro as a more powerful tool, capable of providing solutions when Sonnet fails, especially in more complex or difficult scenarios. “There were a few really cool instances tonight where Sonnet was struggling heavily, and o1 Pro one-shotted a solution,” he said.

Deedy Das, VC at Menlo Ventures, used o1 Pro to solve NYT Connections — a simple game where you group 16 words into four groups of four. “o1 Pro solves it consistently in one shot,” he wrote, on X. 

OpenAI researcher Noam Brown, in a recent podcast, said that he uses o1 Pro for complicated coding tasks. “If I have something that’s pretty easy, I’ll give it to GPT-4o, but if I have something that I know is really hard or that I need to write a lot of code for, I’ll just give it to o1 and have it do the whole thing on its own,” said Brown. 

He said an interesting thing about o1 is that it’s a real proof of concept — users can give it a difficult problem, and it can figure out the intermediate steps on its own and how to tackle those steps. However, he added that he doesn’t really know how people will use o1 until it is deployed in the wild.

Unlike previous models where the chain of thought was primarily a prompting technique, o1 has been trained with reinforcement learning to apply a chain of thought reasoning without additional prompting.  Moreover, OpenAI recently introduced reinforcement fine-tuning (RFT), allowing organisations to create expert-level AI for tasks in law, healthcare, finance, and more. This method enables training with minimal data, sometimes as few as 12 examples.

Unleashes Creativity to a Whole New Level 

OpenAI launched its highly anticipated video-generation model, Sora, which, according to the company, is a calculated step toward achieving AGI one frame at a time. 

With an OpenAI Plus account, users get 50 Sora generations per month, and with a Pro account, users get 500 fast generations (or fewer at high resolution) and unlimited generations in a slower mode. This makes the $200 price tag worthwhile. 

Altman describes the new video-generation model as the ‘GPT-1 for video’ and is eager to see how users will collaborate with each other. “One of the most exciting things to me about this product is how easy it is to co-create with others; it feels like an interesting new thing! This is early – think of it like GPT-1 for video –but I already think the feed is so compelling,” he said.

Sora uses credits for video generation, with costs varying based on video resolution and duration. For example, a 5-second 480p square video costs 20 credits, while the same length in 1080p costs 200 credits. Longer videos incur higher credit charges. Features like Re-cut, Remix, Blend, and Loop also consume credits based on video length. ChatGPT Pro users can generate relaxed videos without using credits.

ChatGPT Plus offers 1,000 credits for up to 50 priority videos (720p, 5s max), while ChatGPT Pro offers 10,000 credits for up to 500 priority videos (1080p, 20s max) and unlimited relaxed videos. 

Back in 2023, when OpenAI launched ChatGPT Professional, it was priced at $42. Notably, a possible explanation for the high price is OpenAI’s anticipated $5 million loss this year.

Is it Worth Buying? 

Many quixotic intellectuals believe the price of ChatGPT Pro is justifiable – $200 is the price of unleashing creative freedom and unlocking scientific discoveries,  among other things.

Ethan Mollick, associate professor at The Wharton School, who got early access to o1, shared his experience and compared it to Claude Sonnet 3.5 and Gemini. “It can solve some PhD-level problems and has clear applications in science, finance, and other high-value fields. Discovering uses will require real R&D efforts,” he said.

He further explained that while o1 outperforms Sonnet in solving specific hard problems that Sonnet struggles with, it doesn’t surpass Sonnet in every area. Sonnet remains stronger in other domains. “o1 is not better as a writer, but is often capable of developing complex plots better than Sonnet because it can plan ahead better,” he said.

A Reddit user shared their experience after spending eight hours testing OpenAI’s o1 Pro ($200) against Claude Sonnet 3.5 ($20) in real-world applications. For complex reasoning, the o1 Pro was the winner, providing slightly better results but taking 20-30 seconds longer per response. 

Claude Sonnet 3.5, while faster, achieved 90% accuracy on these tasks. In code generation, Claude Sonnet 3.5 outperformed o1 Pro, producing cleaner, more maintainable code with better documentation, whereas o1 Pro tended to overengineer solutions.

For advanced mathematics, o1 Pro excelled at PhD-level problems, but Claude Sonnet 3.5 handled 95% of practical math tasks perfectly. In vision analysis, o1 Pro stood out with detailed image interpretation, a capability that Claude Sonnet 3.5 currently lacks. When it came to scientific reasoning, the result was a tie – o1 pro offered deeper analysis, while Claude Sonnet 3.5 provided clearer explanations.

However, Abacus AI chief Bindu Reddy, in her internal tests, pointed out that o1 lags behind Sonnet and Gemini in coding. “We ran the entire live bench AI coding questions by hand and evaluated o1 for coding. The result is that o1 is good but not as good at coding as Gemini or Anthropic. However, it is an improvement over the o1-preview in this category,” she said.

With eight days still left to go of ‘12 Days of OpenAI’ shipmas, ChatGPT Pro users can expect more delights, possibly including advanced voice mode with vision and more.