Liquid AI: US AI Unicorn Enters the German Market via Partner Vago Solutions
US company Liquid AI, one of the younger AI unicorns to emerge from the orbit of the Massachusetts Institute of Technology (MIT), is looking for a way into the German market. As its local partner, the company has brought on Vago Solutions GmbH from Hennef, a firm specialising in tailor-made AI. The stated goal of the collaboration, announced as a “development partnership”, is to build efficient AI models that companies and public authorities can run on their own infrastructure – on-premise rather than in the cloud.
In Austria, there were efforts for some time to get Liquid AI to set up shop locally, given that co-founders Ramin Hasani and Mathias Lechner have their scientific roots at TU Wien – but so far those efforts have not succeeded (more on that here).
Who is Liquid AI?
Liquid AI was founded in 2023 as a spin-off from the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). Its founding team includes Ramin Hasani (CEO), Mathias Lechner (CTO), Alexander Amini and robotics researcher Daniela Rus. The company is based in Brookline near Boston and now employs around 120 people.
Its technical approach differs from that of the major providers such as OpenAI, Anthropic or Google: Liquid AI does not rely on the widely used transformer architecture, but instead on so-called Liquid Neural Networks. The underlying idea traces back to research inspired by the nervous system of the roundworm Caenorhabditis elegans – an organism that is remarkably adaptable with only a few hundred neurons. On this basis, the company develops its Liquid Foundation Models (LFMs), which are optimised for efficiency and are meant to get by with considerably less computing power than conventional large language models.
Liquid AI rose to prominence in late 2024 with a Series A round of 250 million US dollars, led by AMD Ventures – lifting the valuation to around 2.35 billion US dollars and pushing the company into unicorn status. The proximity to chipmaker AMD is strategic: the efficiency-focused models are intended to reduce dependence on powerful, expensive Nvidia GPUs.
Since 2025, Liquid AI has expanded its product range towards edge and on-device applications. With its open-source LFM2 and LFM2.5 models (ranging from 350 million to around 1.2 billion parameters, some with audio and vision variants), as well as its LEAP deployment platform and the Apollo app, the company positions itself as a provider of AI that runs directly on smartphones, laptops, in vehicles or on corporate servers.
What Vago Solutions brings to the table
Vago Solutions is a small, specialised AI development company from Hennef in North Rhine-Westphalia, founded by computer scientists Daryoush Vaziri and David Golchinfar. Within the developer community, the team became known above all for SauerkrautLM, an open-source model geared towards the German language that at times held top spots in relevant Hugging Face rankings.
In the partnership, Vago takes on the task of adapting the Liquid models to specific industries and domains – through fine-tuning and architectural expertise. The aim is to create AI solutions that customers can operate on their own infrastructure without passing sensitive data on to external cloud services.
“Liquid AI thinks about AI from the architecture up – and that’s exactly the level at which we work,” says co-founder Daryoush Vaziri. For a “small, specialised team from Germany”, he adds, a partnership at this level is a validation of the company’s own direction.
Mathias Lechner, CTO and co-founder of Liquid AI, points to the efficiency of the architecture: it is designed “from the ground up” to run on any hardware, he says. In his view, companies should not have to choose between capable AI and control over their own data.
Analysis: Small models, big demand
The partnership taps into a trend that is gaining traction across the German-speaking region: compact, specialised models for individual tasks rather than large general-purpose systems. For public authorities, industry and mid-sized businesses, local deployment is often relevant for data-protection and compliance reasons; added to this are cost and energy arguments, since smaller models require less computing power.
“Most companies don’t need a model that can do everything, but one that masters exactly their task – on their own servers and at predictable costs,” argues Vago co-founder David Golchinfar.
At the same time, the field is competitive: alongside specialised edge providers, the major model houses are also pushing into the same market with on-premise and open models. The partners did not disclose financial details or the scope of the cooperation. In the coming months, Vago says it intends to translate the collaboration into concrete customer projects and give Liquid AI a broader footing in Germany.
