A really interesting claim in the sense that multimodality naturally increases a model’s compute/memory requirements. SEYE says that more useful outcomes are now achieved than before, but compared to what? Multimodality would have been press-release-worthy back in 2023–2024, but nowadays it’s already so common that this press release is quite baffling on many levels 
Surely I must be misunderstanding something here, but AI agents seem like a foolish addition to SEYE’s offering in every way anyway. Some general-purpose classifier/Jev clone would be a significantly more sensible and usable solution if they really have to cram AI into it.