We use cookies to improve your browsing experience. Find out more in our privacy policy.
DeepSeek-R1 – a new era of open-source LLMs
Sophia Antonin Senior UX Data Scientist • Technology Manager AI/ Data Science
Björn Busch-Geertsema Head of Development
Stefan Schulz UX Director • Head of Site Munich
14/02/2025 • 5 minutes reading time
Potentials and Challenges — a summary by Ergosign
Large Language Models (LLMs) are increasingly revolutionizing the digital world. Recently, the Chinese LLM DeepSeek-R1 (and -V3) has made headlines. With unexpectedly high efficiency and performance, it has not only stormed the app charts but also stirred up tech giants and investors. In this two-part Insights article, we will first look at LLMs in general and then delve deeper into DeepSeek.LLMs form the basis for numerous AI applications that come onto the market almost daily. These models are trained over months on huge amounts of text data using machine learning. After this training, they can understand, complete, translate, and even write texts creatively. Well-known examples are ChatGPT, LLaMA, and Claude.But what actually makes such a model, and what should influence the decisions for using certain LLMs in the company? Why are open-source LLMs like DeepSeek so interesting for companies?
A spotlight on Werner
Werner is essentially the mascot of the Fichtelgebirge — both in reality and in his digital version. This creates a high recognition value, even far beyond the borders of the Fichtelgebirge. People know Werner from the app, promotional films, print, and online media, so it was clear that Werner had to be on the kiosk as well.
He acts as a quick start guide, and after the first interaction with the kiosk, the character goes through various states as he shrinks down, but he never disappears completely. Werner is always available and serves as an anchor point for people who are not very familiar with interacting with a digital screen. This simplifies and makes navigation of the application more welcoming. By using Rive for the animation design, the transitions feel particularly smooth, and the movements appear natural.
A spotlight on Rive
Rive is particularly well-suited for animations because it offers endless possibilities for high-detail animations. Additionally, Rive makes the handover to the developers easy. The Rive animation software, like Figma, is primarily a browser application, making it flexible and usable across different devices.
Developers receive a detailed handover with an animation preview, so they can understand exactly how the animation should behave and test it independently. However, design and development work hand in hand with Rive. The design team can do a lot of preparatory work and predefine the states. The development team can initialize the triggers provided by Rive, giving them full control over the animation.
The LLM landscape
Rive X Flutter integration
Both tools are regularly used by us, both separately and in combination. Rive animations frequently find their place on our website as well as in client projects, inviting interaction or creating a 'Wow' effect. Flutter, on the other hand, has established itself as an industry standard in app development.
Flutter allows pixel-perfect development with full control over the appearance of elements. The software is based on a single, cross-platform codebase, allowing the application to run on Android, iOS, and, with minimal adjustments, on desktop and web platforms as well.
Both tools are cross-platform by nature. The widget-based architecture of Flutter and the user-friendly APIs of Rive ensure that animations can be integrated into an app quickly and efficiently.
Rive X Flutter integration
Both tools are regularly used by us, both separately and in combination. Rive animations frequently find their place on our website as well as in client projects, inviting interaction or creating a 'Wow' effect. Flutter, on the other hand, has established itself as an industry standard in app development.
Flutter allows pixel-perfect development with full control over the appearance of elements. The software is based on a single, cross-platform codebase, allowing the application to run on Android, iOS, and, with minimal adjustments, on desktop and web platforms as well.
Both tools are cross-platform by nature. The widget-based architecture of Flutter and the user-friendly APIs of Rive ensure that animations can be integrated into an app quickly and efficiently.
Sophia Antonin hat 2019 ihren Master in Computerlinguistik an der LMU München abgeschlossen. Seither realisiert sie erfolgreich Projekte in den Bereichen Natural Language Processing und Generative KI. Mit ihrer Expertise und Leidenschaft für Künstliche Intelligenz entwickelt sie bei Ergosign innovative Lösungen und gestaltet die digitale Zukunft mit.
Sophia AntoninTechnology Manager AI/ Data Science
Hast du Fragen? Wir freuen uns auf den Austausch.
Die Zukunft der KI wird nicht nur durch leistungsfähige Modelle geprägt, sondern auch durch die Art und Weise, wie sie menschliche Interaktionen, Technologien und Unternehmen transformiert. DeepSeek-R1 ist ein vielversprechender Schritt in diese Richtung.