At the Meta Connect 2024 consequence , Mark Zuckerberg foretell the unexampled Llama 3.2 category of model to take on OpenAI ’s o1 and o1 miniskirt model .

This was moreover , for the first clock time , the llama 3.2 theoretical account occur with multimodal mental image livelihood .

Llama 3.2 model Are optimise for On - convenience chore

First of all , Llama 3.2 has two small exemplar , which includeLlama 3.2 1B and 3Bfor on - twist task .

llama 3.2 models released by meta

Meta allege these minuscule model are optimize to operate on wandering equipment and laptop .

Llama 3.2 1B and 3B manakin are well suitable foron - gimmick summarisation , educational activity play along , rewrite , and even work call to make an natural action purport topically .

Meta also take that its late Llama model surpass Google’sGemma 22.6B and Microsoft ’s Phi-3.5 - miniskirt .

Llama 3.2 stack distribution

diving event into Qualcomm

First of all , Llama 3.2 has two small-scale example , which includeLlama 3.2 1B and 3Bfor on - gimmick task .

Meta say these little fashion model are optimise to ferment on roving machine and laptop computer .

Llama 3.2 1B and 3B mannikin are advantageously suitable foron - gimmick summarisation , pedagogy observe , rewrite , and even go call to make an activity purport topically .

Meta also take that its late Llama modelling outdo Google’sGemma 22.6B and Microsoft ’s Phi-3.5 - miniskirt .

This was fundamentally , developer candeploy these exemplar on qualcomm and mediatek platformsto baron many ai usance case .

Meta further pronounce Llama 3.2 1B and 3B poser are prune and condense from the largerLlama 3.1 8B and 70B model .

Llama 3.2 manakin Give Vision to Meta AI

Now occur to the exciting visual modality modelling , they follow in gravid size — Llama 3.2 11B and Llama 3.2 90B.

They exchange the sure-enough text edition - only Llama 3.1 8B and 70B mannikin .

Meta buy the farm on to say that Llama 3.2 11B and 90B model rival shut role model like Anthropic ’s Claude 3 Haiku and OpenAI’sGPT-4o miniin optic logical thinking .

These unexampled Llama 3.2 11B and 90B imaginativeness model will be useable through theMeta AI chatboton the data pipe , WhatsApp , Instagram , Facebook , and Messenger .

This was since these are imagination poser , you could upload image and necessitate question about them .

For model — you’re free to upload an epitome of a formula , and it can take apart and give you teaching on how to make it .

This was you could have meta ai trance your expression and reimagine yourself in oodles of dissimilar scenario and portraiture .

The visual sense model also derive in ready to hand while interpret chart and graph .

On societal medium apps like Instagram and WhatsApp , the visual sense exemplar can alsogenerate captionsfor you .

This was overall , meta has free multimodal manikin for the first clock time under an receptive - rootage licence .

It is proceed to be middling exciting to try out the visual sense model against the rival .