The German AI consortium has announced Soofi S, an open-source language model containing 30 billion parameters. Soofi S is geared towards industrial AI applications across Europe, integrating a Mixture-of-Experts architecture for efficiency and transparency in model deployment.
The model activates only 3.2 of its 31.6 billion parameters per token, leading to consistent processing speeds, particularly for long inputs. Soofi S's training primarily covered German and English datasets, using Deutsche Telekom's AI cloud infrastructure.
Soofi S outperformed other fully open-source models, such as Olmo 3 32B and Apertus 70B, in benchmarks, excelling in English, German, and coding tasks, despite criticisms over its token-to-parameter ratio surpassing Chinchilla scaling laws.
Initially aimed at technical and regulatory documentation and code-generation applications, Soofi S is poised to be foundational for further specialized models within the Soofi family. It is currently undergoing industry trials with plans for broader availability pending results.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's briefNew every morning, and the back catalogue is archived by date.
The Soofi Consortium has unveiled Soofi S, an open-source model tailored for industrial AI applications in Europe. This 30B Mixture-of-Experts model aims to enhance control, transparency, and efficiency in areas such as regulatory documentation and coding.
The German AI consortium has introduced Soofi S, an open-source language model with 30 billion parameters. It outperforms existing models in English and German benchmarks, emphasizing efficient resource use by activating only 3.2 of its total parameters per token, which ensures constant processing speed.