Farang

What Farang does

Farang is an AI research lab headquartered in Stockholm, Sweden, founded in early 2025 by Emil Romanus—an engineer with over 20 years of experience in search, AI, and startups—following two years of research that yielded a proprietary non-Transformer AI architecture branded as FX. The company's core product is a family of next-generation foundational Large Language Models built on the FX architecture, which is designed from first principles, inspired by biological cognition, and described as comprehending a complete response before generating tokens. Farang claims the architecture delivers 25x greater computational efficiency than Transformer-based models and supports on-device and on-premises deployment for data sovereignty. Initial vertical focus areas include React programming, healthcare and medical applications, and enterprise deployments in regulated industries.

The company operates with a core team of five researchers, developers, and business professionals, supplemented by selective open hiring (including a Head of Developer Experience role). Its go-to-market is a developer-first, community-led motion: early access to FX-powered models is gated through a public waitlist and a request-access form, with planned API access for developers and direct enterprise licensing for on-premises deployments. The intended revenue model combines licensing and royalties (enterprise on-premises deployment) with subscription/recurring access (developer API), though pricing has not been disclosed. Farang is pre-revenue, with use of its €1.5M seed funding (August 2025, co-led by Voima Ventures and Amadeus APEX Technology Fund) directed at scaling proof-of-concept models and acquiring training compute.

Name

Farang

Legal name

Farang

Website

https://farang.com

Company type

Private

Founded year

2025

Operating status

Operating

Headcount range

1–10 employees

Short description

Farang is a Stockholm-based AI research lab developing a proprietary non-Transformer "FX" architecture for specialized Large Language Models, targeting React developers, healthcare, and regulated enterprise customers with on-device deployment for data sovereignty.

Ownership category

akta.pro rank

Product category

AI Foundation Models

NAICS

Software Publishers (5132), Computer Systems Design Services (541512)

SIC

Services-Computer Integrated Systems Design (7373)

akta.pro primary industry

On-Device/Edge Foundation Models (Mobile/Embedded LLMs) (HDAAACAN)

akta.pro secondary industries

Multimodal Generation (Text-Image-Video-Audio) & Creative Toolchains (HDAAACAG), Open-Source Model Ecosystems & Model Marketplaces (HDAAACAM)

Keywords

• Foundational AI models
• Large language models
• AI architecture
• On-device AI
• Enterprise AI deployment

Headquarters

HQ city

Stockholm

HQ country

Sweden

HQ region

Europe

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Infrastructure, Personnel, Marketing or Sales, Operations

Revenue model

1. AI Model Licensing & Deployment: Farang licenses its proprietary FX architecture models to enterprises and organizations for on-premises or on-device deployment. Revenue is generated through model deployment licensing, with particular focus on specialized domain models (programming, healthcare, enterprise) and enterprise-grade deployments with full data sovereignty for regulated industries.
2. Developer API Access: Early access is being granted to a select group of developers, researchers, and partners through a waitlist model, with future plans to offer API access to the FX architecture models. This establishes a consumption or subscription-based revenue stream for developer and researcher access.

Go-to-market motion3 records

Distribution channels4 records

Marketing channels6 records

Core offering

Farang develops and licenses a proprietary next-generation AI architecture called FX that moves beyond the Transformer paradigm. The company builds foundational large language models on this architecture, targeting specialized vertical applications in programming (starting with React), healthcare, and enterprise use cases, with emphasis on on-device and on-premises deployment for data sovereignty.

Product overview

Farang is an AI research lab developing a next-generation AI ecosystem centered on its proprietary FX architecture. The company's portfolio consists of foundational Large Language Models built on this architecture, supplemented by specialized vertical applications targeting programming (starting with React), healthcare, and enterprise use cases. The FX architecture serves as the core technology platform enabling all products, delivering 25x better computational efficiency than Transformers while supporting on-device deployment for privacy-sensitive environments.

Differentiator

Problem solved

Functional benefit

Products and services

• FX Architecture A proprietary next-generation AI architecture developed from first principles that moves beyond the Transformer model. It is inspired by biological cognition, achieves 25x computational efficiency, and enables on-device deployment with privacy by design.
• Farang Large Language Models

Quantifiable outcome

• 25x fewer computational resources required compared to traditional Transformer-based models
• +1 more outcomes

Named customers4 records

Segments3 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability4 records

Feature6 records

Scale indicators4 records

Recent moves5 records

Expansion highlights5 records

Direct peers

• Liquid AI: MIT spinout building Liquid Foundation Models, a non-Transformer architecture with claims of improved efficiency and adaptability. Closely comparable to Farang in approach: novel first-principles architecture, efficiency focus, and enterprise/government commercial path.
• Sakana AI: Tokyo-based AI lab focused on nature-inspired foundation model architectures and efficient training methods. Shares Farang's positioning of building post-Transformer architectures with research-led GTM and significant seed/Series A funding in 2023-2024.
• Mistral AI: Paris-based open-weight foundation model lab serving enterprise and developer markets with efficient open models. Comparable as a European foundation model lab targeting both enterprise licensing and developer adoption, though at materially larger scale.
• Aleph Alpha: Heidelberg-based foundation model company emphasizing European data sovereignty and on-premises deployment for regulated industries. Highly comparable to Farang's regulated-industry, on-prem go-to-market and European AI sovereignty positioning.

Broad incumbents

• AI21 Labs: Tel Aviv-based foundation model company offering enterprise LLMs and task-specific AI applications. Comparable in offering specialized, vertically-tuned enterprise AI products, though operating at larger scale with established customers.
• Cohere: Enterprise-focused foundation model provider targeting regulated industries with private deployment options. Comparable on enterprise/government sales motion and data sovereignty positioning, but serves a broader portfolio of foundation and retrieval models.
• Anthropic: Frontier AI lab offering Claude family LLMs with strong enterprise and API adoption. Serves as the incumbent reference point that Farang is positioning against, particularly in reasoning quality and enterprise safety/sovereignty.
• OpenAI: Incumbent leader in foundation models and ChatGPT enterprise products. Sets the performance and cost baseline Farang must beat, and represents the dominant competitive threat in both developer and enterprise segments.

Emerging players

• Nous Research: Open-source AI research collective shipping fine-tuned open-weight LLMs with strong developer community traction. Comparable on open-science ethos, developer-led distribution, and emphasis on fine-tunable models, though without Farang's proprietary architecture claim.
• Zyphra: Early-stage AI research lab working on novel architectures including state-space model variants. Comparable on novel-architecture research positioning and small-team execution, though earlier-stage and US-based.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat4 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/farang_ai
• GitHubgithub.com/FarangLab
• LinkedInlinkedin.com/company/farang-ai

Revenue estimate

Valuation estimate

Number of profiles

Profiles2 records

Funding overview

Funding rounds1 record

Investors2 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Farang