Originally published on AI Tech Connect.
What shipped: Scout, Maverick, and where to find them Meta released Llama 4 in two open-weight variants designed for different positions on the cost-capability curve. Llama 4 Scout is the efficient, volume-optimised model — lower memory footprint, faster inference, and well-suited to the long tail of requests where throughput matters more than peak capability. Llama 4 Maverick is the higher-capability sibling, carrying more total parameters and aimed at complex reasoning, multi-step analysis, and the harder prompts that Scout would handle less reliably. Both models are available for free download on Hugging Face under the Llama community licence. Meta has also deployed them inside its own consumer products — WhatsApp, Messenger, Instagram Direct, and the Meta AI web interface — which…
Top comments (0)