Back to today’s briefing

Story perspectives

Meta Launches Powerful AI Models, Sparks Evaluation Concerns

4/8/2025

34 5

Summary

Story summary

Meta has unveiled two groundbreaking AI models, Llama 4 Scout and Maverick, boasting superior performance against rivals like OpenAI's GPT-4o. Maverick's impressive ELO score of 1417 on LMArena sparked excitement, but worries emerged over discrepancies between the tested version and the public release. In response, LMArena is set to revise its evaluation policies for fairness.

1 of 1

Story summary

Meta has unveiled two groundbreaking AI models, Llama 4 Scout and Maverick, boasting superior performance against rivals like OpenAI's GPT-4o. Maverick's impressive ELO score of 1417 on LMArena sparked excitement, but worries emerged over discrepancies between the tested version and the public release. In response, LMArena is set to revise its evaluation policies for fairness.