Not just OpenAI! Large language models hit a bottleneck, and Anthropic, backed by Google and Amazon, faces performance stagnation
Anthropic, backed by Google (GOOGL.US) and Amazon (AMZN.US), has also said its large language models have stalled, according to people familiar with the matter.
Three people familiar with the matter said the upcoming Gemini version failed to meet internal expectations. In addition, Anthropic has delayed its next version of its Claude model, Opus 3.5. One of the companies' problems is a lack of human data to train models.
OpenAI CEO Sam Altman has previously said overall compute capacity has been a problem for the latest wave of AI models. Altman recently said on Reddit: "All these models have gotten quite complex and we can't just put out all these things at once." "We're also facing a lot of constraints and hard decisions about how to allocate our compute to all these great ideas."
Anthropic released a new version of its Claude model, Claude3.5Haiku, as well as an upgrade to Claude3.5Sonnet last month.
OpenAI, backed by Microsoft (MSFT.US), has run into performance issues with its upcoming large language model, Orion, according to people familiar with the matter. OpenAI previously denied it would release a model called Orion.
OpenAI, led by Altman, released a series of new artificial intelligence models called o1 and o1-mini in September that can reason through complex tasks and solve problems more difficult than previous scientific, coding and math models.