Llama 4: Did Meta just push the panic button?

a year ago

Meta released Llama 4, which includes three models: Scout, Maverick, and Behemoth, each designed for different use-cases.
The release timing was unusual (on a Saturday) and may have been rushed, leading to speculation about Meta's motives.
Llama 4 introduces mixture of expert architectures, enabling efficient training despite larger total parameters.
The models feature very long context windows (up to 10M tokens) and solid multimodal input performance.
Community reception has been mixed, with criticism over the juvenile behavior of one model and restrictive licensing.
Meta's strategy with Llama 4 seems misaligned with its open-source audience, focusing more on competing with frontier labs.
The release raises questions about Meta's differentiation in AI and its ability to maintain leadership in the open ecosystem.
Regulatory challenges, particularly in the EU, add complexity to Meta's open-source efforts.

Hasty Briefsbeta