Meta's vanilla Maverick AI model ranks below rivals on a popular chat benchmark
This week, Meta got caught red-handed tweaking its Llama 4 Maverick AI to ace the LM Arena benchmark—using an unreleased, chat-optimized version that scored big but wasn’t available to the public. Once exposed, the official version tumbled down to 32nd place, trailing behind older models like GPT-4o and Claude 3.5. Why does this matter for the paper packaging industry? Because AI tools like Llama are increasingly used to optimize packaging design, logistics, and sustainability reporting. If performance data is gamed, packaging companies could make flawed decisions based on overhyped AI capabilities.https://techcrunch.com/2025/04/11/metas-vanilla-maverick-ai-model-ranks-below-rivals-on-a-popular-chat-benchmark/
Comments
Post a Comment