Sber Releases Open Models for Russian AI Stack
At its tenth AI Journey conference in Moscow, Sber announced open-sourcing weights for multiple flagship models, including GigaChat Ultra-Preview and Lightning, Kandinsky 5.0 image/video models, GigaAM-v3 speech recognition and K‑VAE 1.0 compression. The releases include code, documentation and pretrained systems usable by developers. The move aims to integrate Russian-language and multimodal models into global workflows and influence AI standards despite Western sanctions.
Key Points
- 1Publishes open weights for GigaChat Ultra-Preview, Lightning, Kandinsky 5.0, GigaAM-v3 and K-VAE 1.0
- 2Challenges Western containment by contributing production-ready models to the global open-source AI infrastructure
- 3Enables developers and startups to adapt, fine-tune, and embed Russian-language models into products
Scoring Rationale
Significant open-source model release with practical developer impact; limited by primarily Russian-language focus and geopolitical constraints.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Logistics & Shipping problems
