Magibu × Vivax — Native Turkish Multimodal Medical AI Models
— by Vivax
Vivax and Magibu have partnered to build native multimodal medical models in Turkish, to be deployed inside the PreVisit workflow at hospitals where Vivax has…
Vivax and Magibu (magibu.dev) have announced a strategic collaboration to develop a family of native, Turkish-first multimodal medical foundation models. Unlike translated or fine-tuned Western models, these models are trained from the ground up on Turkish clinical language, terminology, and reporting conventions — designed to read text, structured records, images, and voice in a unified clinical context.
The first deployment target is the PreVisit workflow inside hospitals where Vivax has signed formal clinical protocols. PreVisit is the phase before a patient enters the consultation room: history intake, prior-record review, triage signals, and pre-charting. By running natively in Turkish and multimodally across forms, lab results, imaging summaries, and patient voice, the models will compress minutes of pre-visit work into seconds and surface the highest-value context to the clinician.
The collaboration combines Magibu's open model research, training infrastructure, and HuggingFace model releases with Vivax's clinical access, real-world hospital integrations, and PreVisit product surface. Initial model artifacts and checkpoints will be published on Magibu's HuggingFace organization as the program progresses.