For the professional designer, data scientist, or technical illustrator, the promise of generative AI has often ended in a pixelated dead end. A model can conjure a beautiful, photorealistic image of a cell, but the result is a flat picture, not an editable diagram where each component can be resized, recolored, or annotated. This gap between AI-generated imagery and usable, precise vector graphics is the problem Berlin-based F13 is attempting to solve with a new class of foundation model [Tech.eu, September 2026].
Launched from stealth in September 2026 with a $5 million pre-seed round, the company is training models to natively output vector graphics from text, images, or existing graphics [Tech.eu, September 2026]. The goal is not just another image generator, but a tool that produces SVG-style outputs where proportions, layouts, and editability are preserved, addressing what one observer called the failure of existing solutions to generate "usable SVGs" [Lorenz Hufe - Fraunhofer Heinrich Hertz Institute HHI | LinkedIn, retrieved 2026].
The bet on vector-native generation
The company's technical wedge is its focus on the vector format itself. While most popular image models output raster images (grids of pixels), F13's models are designed to generate vector graphics, which are defined by mathematical paths and shapes. This makes them infinitely scalable without quality loss and, crucially, editable in tools like Adobe Illustrator or open-source alternatives.
F13 cites professional workflows requiring precision and consistency as its primary targets. These include generating data visualizations, scientific diagrams, maps, and brand-consistent marketing documents [EU-Startups / EUVC, September 2026]. For a researcher needing a schematic of a protein interaction or a brand manager requiring a set of on-brand icons, an AI that outputs a messy, uneditable PNG is of limited utility. F13 is betting that by training on vector data and the relationships between graphical elements, its models can deliver "real utility" where others have not [Tomas Fanta, CFA - Venture investor, CEE&Baltics, AI, Blockchain | LinkedIn, retrieved 2026]. The company is offering early access through a developer API and a web platform [EU-Startups / EUVC, September 2026].
The team behind the geometry
The founding duo brings a blend of entrepreneurial and mathematical rigor. Gregory Janik, the CEO, is described as a two-time exited entrepreneur who previously worked at the edtech company Knownunity [daily.dev, retrieved 2026] [The Next Web, September 2026]. His co-founder, Ingimar Tomasson, is a machine-learning researcher and mathematician trained at University College London [The Next Web, September 2026]. The two met while working at Knownunity, where they reportedly collaborated on product visuals, a experience that likely informed their focus on the practical shortcomings of existing design tools [cfotech.co.uk, October 2026].
Reports also indicate the broader team has been built with specific expertise in both high-performance AI and high-end design. The team reportedly includes former machine learning researchers from Meta and Porsche, as well as designers who have worked with luxury brands like Versace, Gucci, and Nike [ETIH EdTech News, retrieved 2026]. This combination suggests an understanding of both the computational challenge of training frontier models and the nuanced demands of professional visual communication.
Why Credo and Point Nine wrote the check
The $5 million pre-seed round, led by Credo Ventures and Point Nine Capital with angel participation, is a significant vote of confidence for a company just emerging from stealth [Tech.eu, September 2026]. The investment signals that experienced European early-stage funds see a tangible market need that existing AI image generators are not addressing.
The capital is reportedly earmarked for expanding compute resources, talent, and data,the essential fuel for training frontier AI models [Vestbee, September 2026]. For investors, the bet is that F13 can establish a technical moat by building a proprietary dataset of vector graphics and relationships that general-purpose models lack. If successful, it could own a new category of "precision graphics AI" used by enterprises and developers who need more than just a pretty picture.
Pre-seed | 5 | M USD
The competitive and technical landscape
F13 enters a field crowded with image-generation giants, but its focus on a specific output format and professional use case is its differentiator. The primary competition is not necessarily other startups, but the incumbent workflow: professionals manually creating vectors in design software or settling for inferior raster outputs from AI.
The technical risks, however, are substantial. Training a foundation model capable of understanding and generating complex, hierarchical vector graphics is a formidable challenge. The model must infer logical structure from a prompt,understanding that a "bar chart" has axes, labels, and data series that are semantically connected and editable as separate objects.
Key questions the company must answer in the coming months include:
- Model fidelity. Can the AI consistently produce graphics that are not just visually plausible but factually and dimensionally accurate for technical use cases?
- Editability depth. Does the generated SVG contain logically grouped and labeled elements, or is it a monolithic shape that defeats the purpose?
- Market timing. Will developers and enterprises adopt a specialized API for vector generation, or will they wait for larger multimodal models to incorporate similar capabilities?
The next twelve months for F13
The immediate roadmap involves moving from early access to a broader public launch, which reports indicated was planned for later in 2026 [EU-Startups / EUVC, September 2026]. Traction will be measured by developer adoption of its API and the sophistication of graphics its web platform can produce. The company will also need to demonstrate its models on increasingly complex prompts, moving from simple icons to detailed scientific diagrams or interactive data visualizations.
The patient population here is not defined by a disease, but by a professional constraint: anyone whose work depends on precise, scalable, and editable visual communication. This includes academic researchers, business analysts, cartographers, and brand designers. The standard of care today is a time-intensive manual process in software like Adobe Illustrator or a compromise with a static, uneditable image from a tool like Midjourney. F13 is betting that an AI fluent in the language of vectors can fundamentally alter that workflow, turning a text description into a professional-grade, malleable asset. The next year will show if its models have learned that language well enough to become a staple in the professional toolkit.
Sources
- [Tech.eu, September 2026] F13 emerges from stealth with $5M for vector graphics AI | https://tech.eu/2026/09/23/f13-emerges-from-stealth-with-5m-for-vector-graphics-ai/
- [The Next Web, September 2026] F13 raises $5M pre-seed to build an AI model for precise vector graphics | https://www.thenextweb.com/news/f13-raises-5m-pre-seed-vector-graphics-ai
- [EU-Startups / EUVC, September 2026] F13 raises $5m to build precise vector-graphics AI | https://www.eu.vc/p/f13-raises-5m-to-build-precise-vector-graphics-ai
- [Vestbee, September 2026] F13 launches from stealth with $5M to tackle AI’s visual accuracy gap | https://www.vestbee.com/insights/articles/f13-launches-from-stealth-with-5-m
- [cfotech.co.uk, October 2026] F13 raises $5m to build precise vector-graphics AI | https://cfotech.co.uk/story/f13-raises-5m-to-build-precise-vector-graphics-ai
- [ETIH EdTech News, retrieved 2026] F13 raises $5m to build precise vector-graphics AI | https://etih.io/2026/09/23/f13-raises-5m-to-build-precise-vector-graphics-ai/
- [daily.dev, retrieved 2026] F13 raises $5M pre-seed to build an AI model for precise vector graphics | https://daily.dev/blog/f13-raises-5m-pre-seed-to-build-an-ai-model-for-precise-vector-graphics
- [Lorenz Hufe - Fraunhofer Heinrich Hertz Institute HHI | LinkedIn, retrieved 2026] Post on F13 | https://www.linkedin.com/in/lorenz-hufe/
- [Tomas Fanta, CFA - Venture investor, CEE&Baltics, AI, Blockchain | LinkedIn, retrieved 2026] Post on F13 | https://www.linkedin.com/in/tomas-fanta/