Four divisions.
One architecture.

Crassus is our applied intelligence and research division, operating two distinct terminals on a shared engineering foundation. The Crassus National Intelligence Terminal aggregates real-time economic, commodity, news, and climate data for any of 250 countries - a mobile-first PWA built for governments, sovereign wealth funds, and institutional clients with no per-seat licensing.

The Crassus V5.0 Financial Terminal is the analytics engine behind our published institutional credit research, integrating macro economics (29 FRED series), equity fundamentals, regulatory filings, technical analysis, and options intelligence into a unified analytical environment with 25+ API endpoints.

By deploying AENEA models into live intelligence and financial environments, Crassus ensures our core AI stack remains factually deterministic under the most demanding analytical conditions.

250 countries · 14 commodity markets · 11 published reports · Both terminals live

Visit Crassus.info

Quartz is the foundational layer of AENEA Global. We operate the pipelines that ingest, sanitise, and validate the raw data required to train deterministic language models.

The current flagship QT V.4.4 32K UltraLingo tokenizer covers 72 languages across 27 writing systems, achieving the highest cross-lingual equity ratio of any QT generation while operating at one-quarter the vocabulary size of comparable models. The V.4.1 64K UltraLingo serves as the companion tokenizer for our larger 500M-2B parameter models in the Overture series.

Quartz is AENEA Global's open-source division. All Quartz data pipelines, cleaning scripts, and tokenizer families are published under Apache 2.0 or MIT licences. This open-source commitment is exclusive to Quartz and does not extend to AENEA model weights or Crassus research systems.

Open source · Apache 2.0 / MIT · 72 languages · 27 writing systems

Visit Quartz.host

AENEA designs the model architectures that sit on top of the Quartz data stack. We build small, dense, highly efficient networks designed specifically for factual recall and multilingual reasoning.

Our Prelude model family has demonstrated that when training geometry is aligned with ultra-clean data, parameter counts can be reduced by orders of magnitude without sacrificing reasoning capability. Prelude-5 (276M parameters) is currently in active training on the QT V.4.4 32K UltraLingo tokenizer, reaching the factual crystallisation gradient norm zone five times faster than prior generations - validating our Factual Crystallisation Hypothesis, the framework identifying gradient norm rather than loss as the primary predictor of factual recall emergence in language models.

The Overture model family - our planned next-generation 500M-2B parameter architecture - sits on the QT V.4.1 64K UltraLingo companion tokenizer and builds upon the training insights established by the Prelude series.

Visit AENEA.app

RoamFree is AENEA Global's consumer-facing commercial venture: a UK-domiciled travel eSIM service that gives travellers instant mobile data without roaming fees or a physical SIM swap. Customers browse, purchase, and install a data eSIM in minutes, with coverage spanning 30+ countries and regions and transparent pricing in GBP.

Where the rest of the group builds intelligence infrastructure, RoamFree applies the same operational discipline to a high-volume consumer market. The platform launched in May 2026 and recorded its first live customer sale within days.

RoamFree also runs a self-serve affiliate programme, allowing partners to earn commission on referred eSIM purchases - extending the venture's reach beyond direct sales.

UK-domiciled · GBP pricing · 30+ countries & regions · Affiliate programme live

Visit RoamFree.me