OpenAI CEO Sam Altman said the next advanced version of its GPT large language model series will be coming in “weeks or months,” according to a post he shared on X.

Altman said the next artificial intelligence (AI) model the company will ship is GPT-4.5. It was internally nicknamed “Orion” and will be OpenAI’s last non-reasoning model.

Altman also said OpenAI plans to better organize its AI models. “We realize how complicated our model and product offerings have gotten,” he wrote.

Currently, the startup has two groups of AI models: Its GPT series ranging from GPT-1 to GPT-4, which are non-reasoning models, and its omni series, o1 and o3, which reason. A preview of o1 launched last September. (OpenAI skipped o2 because it’s also the name of a British telecom provider.)

The GPT models generate answers immediately but the reasoning models take time to think. The goal of reasoning models is to come up with more accurate and insightful answers by following a chain-of-thought process.

But with two model families and iterations within each, it was getting too complicated. “We want AI to ‘just work’ for you,” Altman said.

After shipping GPT-4.5, OpenAI plans to unify both GPT and omni series “by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a very wide range of tasks,” he said.

To that end, when OpenAI eventually releases GPT-5, it will integrate both the capabilities of the reasoning and non-reasoning models, Altman said. That also means the o3 model will no longer be a standalone model.

While organizing the product line up “makes a lot of sense,” for researchers and developers, “having everything encapsulated moves us away from being able to understand how things work underneath the hood,” Percy Liang, associate professor in computer science at Stanford University, wrote in an X post.

Altman said free ChatGPT users will get unlimited chats using GPT-5 at the “standard intelligence” setting, he said. Currently, free users only get access to the less advanced GPT-3.5. ChatGPT Plus subscribers ($20 a month) will be able to use GPT-5 at a higher level of intelligence while Pro members ($200 a month) get an even higher level of intelligent capability.

Anthropic Launches ‘Economic Index’

OpenAI rival Anthropic is wading into economics with the launch of an “Economic Index.”

“In coming years, AI systems will have a major impact on the ways people work,” the Amazon- and Google-backed startup said in a blog post. That is why it launched the index to understand “AI’s effects on labor markets and the economy over time.”

For its first report, Anthropic scoured more than 4 million conversations on its AI chatbot, Claude, using a “privacy-preserving” system, according to its researchers. These chats were filtered through the lens of tasks and occupations in a U.S. Department of Labor database.

Highlights of the findings:

• Nearly half of total AI use is for software development and writing.
• Across industries, around 36% of occupations use AI for at least a quarter of their tasks while 4% use it for three-quarters of their tasks.
• 57% use AI to augment human skills and 43% to automate tasks, according to the report.
• AI use is more prevalent for tasks done by mid- to high-wage jobs (computer programmers and data scientists, for example), but lower for both the lowest- and highest-paid occupations.

Anthropic said it is open sourcing the dataset used for the analysis, which would let others improve and add to the research. It also invited economists and policy experts to weigh in due to AI’s impact on jobs and productivity.

NTT Data Debuts New AI Agent

NTT Data, a division of Japanese telecom giant NTT Group, is seeking to generate $2 billion of revenue from its new AI agent by 2027.

Smart AI Agent automatically identifies, organizes, and completes tasks based on user instructions. In doing so, the AI agent “empowers businesses to quickly assess and implement generative AI applications, streamlining operations and driving higher productivity,” the company said.

The AI agent is already enhancing data analysis efficiency in the automotive industry, improving regulatory reporting in the banking industry and optimizing the marketing cycle for a major gas company, NTT Data said.

NTT Data said its AI agent’s capabilities include the following:

• Task planning: Autonomously divides tasks into streamlined processes.
• Multi-agent collaboration: Enables multiple AI agents to work together on workflows.
• Advanced retrieval-augmented generation (RAG): Provides contextual searches of internal corporate data for higher-quality results.
• Agent ops: Generates validation data from business documents to optimize operational processes.
• User-in-the-loop capability: Coming in March 2025, this feature will continuously improve agent workflows based on user feedback.

NTT Data said it serves 75% of the Fortune Global 100 companies.

AI Models: OpenAI Unveils AI Model Plans, Anthropic Launches Economic Index