Top 66%
dans ce catalogue (794 pratiques notées)
Les scores se concentrent dans le haut de l’échelle ; la position au sein du catalogue est donc souvent plus parlante que le chiffre seul.
Evidence of impact / measured public value1/3
Only capability benchmarks (top-ranked Arabic LLM in its parameter class) are public; no independent data on production accuracy, uptake, or citizen/civil-servant impact has been released six months after launch.
Transparency, fairness & accountability2/3
Karnak-40B's weights were released publicly and specific downstream applications (SIA, AcQua, Torgoman, BelMasry) are named, but no bias, fairness or safety audit has been published.
Transferability / demonstrated replication2/3
The same base model has been replicated across at least five distinct government applications spanning education, legal aid, translation, call-centre QA and healthcare, though no adoption outside Egypt is documented.
Scalability beyond pilot1/3
Multiple flagship applications were announced at launch, but the model has been in production only since February 2026 with no reported rollout or adoption figures across Egypt's public institutions.
Governance, capability & sustainability2/3
Karnak sits within Egypt's National AI Strategy (2025-2030) and was launched by MCIT with presidential-level visibility, indicating institutional backing, though sustained funding and long-term operating structure are not yet documented.
Evidoria. Karnak — le Grand Modèle de Langue Arabe Souverain de l'Égypte au Service des Applications Gouvernementales d'IA. ID persistant: d03c0288-df93-439e-a732-53bcfdec8bd1.