Meta Llama 4 Accused of "Cheating": Aces Arena Tests but Fumbles in Real-World Use
Meta Llama 4 AI controversy: Top benchmark scores but real-world coding struggles. Is it benchmark-optimized?
"AI Disruption" Publication 5700 Subscriptions 20% Discount Offer Link.
Meta’s stumble came out of nowhere.
Last Saturday, Meta released its latest AI model series—Llama 4—and dropped three versions at once: Llama 4 Scout, Llama 4 Maverick, and Llama 4 Behemoth.
According to the official introduction, their rankings in the large model arena are pretty impressive.
Take Llama 4 Maverick, for example—it’s ranked second overall, becoming the fourth model to break the 1400-point barrier.