15
Badanie BCG wśród 940 marketerów i specjalistów agencji influencer marketingowych; tylko 15% respondentów ma dojrzały model operacyjny
BCG2026global
A family of open-source language models from Meta. The release of model weights in 2023 triggered a wave of open models, providing a starting point previously available only to large companies.
Why it matters
It is the most commonly used open model foundation. A significant portion of what runs locally on regular hardware today is either based on Llama or closely resembles it.
What's missing without it
Without open weights, working with language models reduces to using a provider’s API — subject to their pricing, limits, and visibility into submitted content.
When it is used
In locally run models and as a component in larger systems: in the HiDream image generator, it’s Llama that reads the scene description.
How we use it
Meta’s license requires acceptance of terms upon download. Community-published versions, such as those prepared by Unsloth, are available without this barrier.
Numbers worth knowing
LLAMA in data
40%
Nowy model Claude Opus 5.5 ma kosztować 40% mniej niż Opus 5 przy typowych zadaniach
Anthropic09/2026global
10%
Model 10-20-70 BCG: 10% algorytmy, 20% technologia i dane, 70% ludzie, organizacja, procesy
BCG2026global
3
model opracowany przez Faculty (firmę AI przejętą przez Accenture) dla NHS England; AI wspierała decyzje, ale odpowiedzialność pozostawała po stronie liderów NHS
Accenture09/2026Europe