Skip to main content

f-inference

HORIZON EUROPE · 2026—2029

AI intelligence,

where it matters.

Foundation model inference, efficient from cloud to edge to device.

f-inference is developing an open framework to make foundation models faster, more energy-efficient and affordable, while keeping data under the control of those who own it.

Cloud

HIGH-CAPACITY COMPUTE
Scalable resources when needed

Edge

LOCAL EXPERTISE
Adaptive inference close to data

IoT

REAL-WORLD DEVICES
Sensors, vehicles and embedded systems

36 months

RESEARCH & INNOVATION

19

PROJECT PARTNERS

12

COUNTRIES

3

PILOT DOMAINS

€7.98 M

EU CONTRIBUTION (ROUNDED)

MEET F-INFERENCE

Reimagining where AI runs.

Foundation models are reshaping modern technology. But relying only on centralised clouds can create high costs, longer response times and concerns over data sovereignty.

Our three-year research action brings together universities, research centres and industry to distribute intelligent inference across a Resource-driven Computing Continuum.

About the project →

CLOUD ↔ EDGE ↔ IoT

One coordinated continuum. The most suitable model, placed on the most suitable resource, at the most suitable time.

WHY IT MATTERS

Better inference.
Better outcomes.

Four challenges that drive the f-inference research agenda.

01 / DIGITAL SOVEREIGNTY

Keep data in your control

Run models where sensitive data is generated, reducing reliance on external infrastructure.

Explore

02 / ENERGY EFFICIENCY

Smaller footprint. Smarter inference.

Reduce energy through model optimisation and resource-aware placement.

Explore

03 / INCLUSIVE ACCESS TO AI

Powerful AI for more organisations

Bring efficient foundation models within reach of SMEs and resource-limited settings.

Explore

04 / LOW LATENCY

Real-time intelligence

Move inference closer to people, machines and sensors for responsive applications.

Explore

OUR METHODOLOGY

From optimisation
to orchestration.

Three connected steps make the computing continuum practical.

Optimise

Lighter models suited to each device: profiling, compression, early exit, collaborative training.

Execute adaptively

Split, route and cache inference so each request reaches the most suitable model or node.

Orchestrate & validate

The open f-inference Framework deploys models against cost, energy and sovereignty goals.

REAL-WORLD VALIDATION

Three domains, one framework.

Three industrial-led pilots will test the approach in challenging environments.

Digital industry

Foundation models that help manage the computing and network resources of an industrial plant.

Automotive

Situational awareness for connected and automated vehicles, with privacy visible at every tier.

Healthcare

Predictive maintenance for reactors producing medical radioisotopes.

19 partners from 12 countries
LATEST FROM THE PROJECT

New research.
Shared progress.

Follow advances in distributed AI, project milestones and new opportunities to connect with the f-inference ecosystem.

UPCOMING · 26–27 OCTOBER 2026

f-inference project kick-off in Athens

Partners are scheduled to meet to launch the research programme and align on the three pilot domains.

Visit News & Events →

STAY IN THE LOOP

Insights worth sharing.

Get updates about events, publications, pilots and open project results.

CONNECT WITH F-INFERENCE

Help shape the next generation of efficient AI.

Explore collaboration opportunities, research activities and real-world pilots.