SoofiS Releases Milestone Results and Model Weights
mishig25 · x · 2026-07-13
A consortium of German institutions and companies has released the results of the SoofiS project, providing the report and all artifacts.
- Participants include DFKI, Fraunhofer IAIS / IIS, TU Darmstadt, University of Würzburg, BHT, L3S, Lamarr Institute, ellamindAI, Hessian AI, and Merantix, coordinated by ki-verbands and funded by BMWE / IPCEI-CIS / 8ra.
- The project's model weights are currently in a gated beta on Hugging Face and will be freely available once the beta ends.
- The author mentions this was a training run on the scale of 27 trillion tokens, with the team continuously troubleshooting during nights and weekends to ensure cluster stability.
The post also includes links to the report and related resources, reading overall like a milestone release for a sovereign AI / open-source LLM project.
Related event: German Alliance Releases Open-Source Sovereign Model Soofi S 30B-A3B(5 posts)→
More from Models
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Claude is no longer available for minors as Anthropic rolls out age assurance — Muhammad523 · 2026-09-11
- Developer Building a Unified Leaderboard of All Model Benchmark Scores — airesearch12 · 2026-09-11
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11