EUWL

Intelligence in orbit

SCROLL TO EXPLORE

Our Mission

We orbit the hardest problems in artificial intelligence.

01

Fundamental Research

Pushing boundaries in language understanding, reasoning, and multi-modal intelligence at the frontier of what machines can comprehend.

02

Safety & Alignment

Ensuring advanced AI systems remain interpretable, controllable, and aligned with human values as capabilities scale beyond current thresholds.

03

Real-World Systems

Translating breakthroughs into deployable systems that augment human decision-making across science, medicine, and industry.

Research Areas

The Gravity Well

Core Capability

General Language Models

Our flagship research path — models that understand, reason, and communicate across every domain of human knowledge with unprecedented depth.

47T

Parameters in latest training run

Safety

Alignment Research

RLHF, constitutional AI, and novel reward modeling techniques.

98.4%

Alignment score

Multimodal

Vision-Language

Unified architectures that see, read, and understand simultaneously.

12

Modalities integrated

Infrastructure

Distributed Training

Custom clusters across Europe running our proprietary distributed training framework for frontier-scale models.

3.2 EF

Peak compute

Publications

Constellation

Our publications form a map — each one a star, connected by the gravity of shared inquiry.

2024

Scalable Alignment via Constitutional Self-Play

2024

Unified Multimodal Transformers for Cross-Domain Reasoning

2023

Efficient Sparse Attention for Long-Context Understanding

2024

Emergent Planning in Autoregressive Models

2023

Formal Verification of Neural Network Behaviour Under Distribution Shift

2023

Quantum-Classical Hybrid Training for Edge Deployment

Investment Focus

Spectrum

How we distribute our research energy across the electromagnetic spectrum of intelligence.

Foundation Models

92%

Safety & Alignment

78%

Multimodal AI

65%

Applications

51%

Deep Space Signal

Transmission

SIGNAL ACTIVE
FREQ: 2.47 GHz

[ORIGIN] EUWL LAB — ZURICH
[TIMESTAMP] 2024.12.07 — 03:47:22 UTC
[CONTENT] BREAKTHROUGH DETECTED
[STATUS] VERIFIED — PUBLISHING

"We observed genuine chain-of-thought reasoning emerge at scale — not pattern matching, but structured deduction our models had never been explicitly trained to perform."

Begin

Begin your
orbit

Join the trajectory of Europe's most ambitious AI research laboratory. Every great discovery begins with a single point of entry.