Google, Anthropic also run into issues with plateauing model performance: report

Like OpenAI, Google (NASDAQ:GOOG) and Amazon (NASDAQ:AMZN) backed Anthropic are also said to be running into plateauing performance of their large language models, according to Bloomberg.

An upcoming version of Gemini is not living up to internal expectations, the news outlet reported, citing three people with knowledge of the matter. Additionally, Anthropic has delayed the next version of its Claude model, known as 3.5 Opus. One of the issues that the companies are running into include a lack of human-made data to train the models on.

OpenAI CEO Sam Altman has said in the past that overall computing power is also an issue for the latest wave of AI models. “All of these models have gotten quite complex, and we can’t ship as many things in parallel as we’d like to,” Altman said on Reddit recently. “We also face a lot of limitations and hard decisions about how we allocate our compute towards many great ideas.”

Anthropic announced a new model, known as Claude 3.5 Haiku, along with an upgraded version of Claude 3.5 Sonnet, last month.

Google and Anthropic did not immediately respond to a request for comment from Seeking Alpha.

OpenAI, which is backed by Microsoft (NASDAQ:MSFT), has reportedly run into performance issues of its own on its upcoming large language model, code-named Orion. OpenAI has refuted in the past that it will release a mode known as Orion.

The Altman-led OpenAI unveiled a new series of AI models in September, known as o1 and o1-mini, that can reason through complex tasks and solve harder problems than previous models in science, coding, and math.