Meta executives were focused on rivaling OpenAI’s GPT-4 during the development of its Llama 3 AI model, according to court filings unsealed on Tuesday, in the lawsuit Kadrey v. Meta.

Internal messages from 2023 and 2024 reveal that Meta’s AI team set GPT-4 as their benchmark, along with Anthropic’s Claude.

Ahmad Al-Dahle, Meta’s VP of generative AI, emphasized the need to secure significant resources, including 64,000 GPUs, to compete with OpenAI and Anthropic.