ref=”dofollow” data-ga-onclick=”Inarticle articleshow link click#Tech#href” href=”https://economictimes.indiatimes.com/topic/meta” target=”_blank”>Meta’s newest open-weight large language model, Llama 4, is now out — aimed directly at developers, ref=”dofollow” data-ga-onclick=”Inarticle articleshow link click#Tech#href” href=”https://economictimes.indiatimes.com/topic/ai-startups” target=”_blank”>AI startups, and researchers looking for alternatives to OpenAI’s GPT-4 and Google’s Gemini.
Available in two variants, Scout and Maverick, the model uses a mixture-of-experts (MoE) architecture that activates only parts of the model per query, making it more efficient to run. Here’s how to start using it.
Where to get it
data-ga-onclick=”Inarticle articleshow link click#Tech#href” href=”https://economictimes.indiatimes.com/tech/technology/meta-releases-new-ai-model-llama-4/articleshow/120031176.cms?from=mdr” data-type=”tilCustomLink” target=”_blank”>Meta has made Llama 4 available via its GitHub repository, with access gated through a request form for responsible usage….
