LMArena published more than 2000 comparisons of the Meta* model 'Llama-4-Maverick-03-26-Experimental', which alleviated concerns about its features. The platform stated: 'Meta*'s interpretation of our policy does not align with what we expect from model providers'. Data shows that Llama 4 generates longer and formatted responses with frequent use of emojis. Artificial Analysis updated the scores for Llama 4 after reviewing evaluation criteria: the Scout intelligence index increased from 36 to 43, while Maverick rose from 49 to 50 in MMLU Pro and GPQA Diamond tests. The 'standard' Maverick model ranks 32nd on the LMArena leaderboard.