Key players include Meta Platforms, led by CEO Mark Zuckerberg who announced it via Threads, and Meta Superintelligence Labs. The development addresses competition from OpenAI, Anthropic, Google (Gemini), and others like DeepSeek and xAI (Grok). Models like Llama 4 Scout, Llama 4 Maverick (17B active parameters, competitive with GPT-4o on benchmarks), and previewed Llama 4 Behemoth highlight Meta's push.[1][2][3][4]
This follows disappointment from Meta's last major AI release over a year ago, prompting a billions-dollar overhaul in hiring and operations to catch up. The new models outperform predecessors like Llama 3 on multilingual support (10x more tokens), coding, reasoning, and bias reduction, marking Meta's first significant update in over a year.[1][2][3]
Newsworthy now as it tests Meta's AI ambitions amid fierce rivalry, with Muse/Spark rivaling top models on benchmarks while introducing cost-efficient, high-performance features like 400B total parameters in Maverick at lower prices than prior versions.[2][3][4]