OpenAI’s AI-powered chatbot platform, ChatGPT, now has a new, very expensive membership plan.
Confirming rumors that came out this morning, OpenAI announced ChatGPT Pro, a new $200-a-month plan that gives users full access to all of OpenAI’s models, including the o1 “reasoning” model.
“Power users of ChatGPT will be interested in ChatGPT Pro,” said Jason Wei, a member of OpenAI’s technical staff, during a livestreamed press conference on Thursday. “These are the people who are already pushing the models to the limits of their abilities in tasks like math, programming, and writing.”
o1 and other thinking models try to check their own work as they go, which is different from most AI. While this helps them avoid some of the problems that models often run into, it means that they take longer to find answers. O1 thinks through tasks, makes plans, and does a number of things that help the model figure out what to do.
A preview of o1 was released by OpenAI in September, but this new version is typically faster. What users can look forward to is “a faster, more powerful, and accurate reasoning model that is even better at coding and math” than the preview, an OpenAI representative told Parhlo World.
o1 can now think about picture uploads, which it couldn’t do during the preview. To improve response times, it has been trained to be “more concise in its thinking.” OpenAI’s own tests show that o1 cuts down on “major errors” on “difficult real-world questions” by 34% compared to the preview version.
But for some reason, the full o1 version does worse on a number of popular benchmarks than the preview version. MLE-Bench is one of these standards. It checks how well AI “agents” do at machine learning engineering.
O1 doesn’t need to be a ChatGPT Pro member. Through the ChatGPT model selection tool, all paid ChatGPT users can now get to o1 as of this afternoon.
But people who pay for ChatGPT Pro will supposedly get a better version of o1 than people who don’t pay as much. They say it “uses more compute for the best answers to the hardest questions” and it’s called o1 pro mode.
ChatGPT Pro users can use the feature by going to the model picker and choosing “o1 pro mode.” They can then ask a question directly. ChatGPT will show a progress bar and send an in-app notification if they switch to another chat while answers are being made.
In O1 pro mode, the model may just take longer to “reason” before giving an answer. The o1 preview statement from OpenAI said that the company wanted to test o1 models that can reason for hours, days, or even weeks to make them smarter. This could be a step in the right direction.
An OpenAI representative told Parhlo World that feedback from outside expert testers shows that o1 pro mode consistently gives more complete and correct answers, especially in areas like data science, programming, and case law analysis. ” o1 pro mode does better on difficult machine learning tests in math, science, and coding than both o1 and o1-preview.” We saw a 75% drop in mistakes on questions from the easier coding challenge that were more like real-life programming problems.
Soon, OpenAI’s API will also have O1. This will add new features like function calling (so you can use outside tools) and picture analysis. OpenAI says that in the coming months, it will add things like web viewing, file uploading, and more.
OpenAI’s most expensive plan so far is ChatGPT Pro, which costs 10 times as much as ChatGPT Plus. Only the most loyal users are likely to buy it, since a lot of people already think ChatGPT Plus is too expensive.
To make things even better, ChatGPT Pro gives you endless access to GPT-4o and ChatGPT’s human-like conversational mode, Advanced Voice Mode.
They will also give away some free membership. The company announced a program to give 10 grants of ChatGPT Pro to medical researchers at “leading institutions.” In the future, the company plans to give more grants in “various disciplines.”
For a long time, people have said that the special ChatGPT’s price would go up. A story in The New York Times says that OpenAI plans to charge $44 a month for ChatGPT Plus by 2029. According to The Information, the company has also thought about offering very expensive business packages with extra features and access to models that are still being worked on. The news of today certainly backs up those stories.
The bold moves show that investors are putting pressure on OpenAI to cut its losses. According to The New York Times, the company made $300 million a month in August. However, OpenAI plans to lose about $5 billion this year. Staffing, office rent, and building up AI training systems are some of the costs that are to blame. At one point, OpenAI was said to be spending $700,000 a day on ChatGPT alone.
ChatGPT is still one of OpenAI’s main ways to make money. Over 300 million people use the site every week, and about 10 million of them pay to be a subscriber.
What do you say about this story? Visit Parhlo World For more.
