New AI Models Coming Soon From OpenAI and Meta

According to a Financial Times report, Open AI’s GPT-5 is reportedly due for release this summer. Additionally, Meta confirmed that it will launch an early version of Llama 3, the next generation of its large language model (LLM), next month.

Early testers who spoke with Business Insider said the new model is “materially better” than GPT-4. While many consider GPT-4 one of the most advanced – and most popular – generative AI technologies at the moment, OpenAI’s CEO Sam Altman disagrees. He slammed it on a recent podcast with Lex Fridman, saying, “I think [GPT-4] kind of sucks,” and added that OpenAI “will release an amazing new model this year.”

No word from Altman on what “amazing” means in technical terms, but both OpenAI and Meta are working on models capable of reasoning and planning. Instead of just generating text based on patterns, the upcoming large language models will be capable of logical thinking, planning, executing multi-stage tasks, and remembering.

GPT-5 can reportedly interact with other AI programs developed by OpenAI, known as “AI Agents,” that can perform tasks autonomously. This means AI systems within the OpenAI ecosystem can collaborate on completing complex tasks on their own.

Before GPT-5 sees the light of day, the public will get access to Meta’s new suite of next-generation LLMs. According to Nick Clegg, the company’s President of Global Affairs, the Llama 3 model will power multiple Meta products. “There will be a number of different models with different capabilities, different versatilities [released] during the course of this year, starting really very soon.”

Users of Meta’s previous versions of Llama deemed it too limited, with the general public preferring Microsoft-backed OpenAI. Despite its slower progress, Meta has big plans for its upcoming LLM. “Our goal over time is to make a Llama-powered Meta AI be the most useful assistant in the world,” said Joelle Pineau, VP of AI Research.

While “there’s quite a bit of work remaining to get there,” Llama 3 should at least answer questions more accurately and possibly respond to both text and image prompts. Also, as an open-source model, it’s expected to gain a more favorable position among developers.