Locai Labs
Locai Labs is a London-based British foundational AI company that builds sovereign large language models and on-premises AI hardware, post-training open base models on customer data so regulated enterprises and governments can own the model, weights, and IP outright.
- Company typePrivate
- Founded2022
- HeadquartersLondon, United Kingdom
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Locai Labs does
Locai Labs Limited is a privately held British foundational AI company headquartered in London (Michelin House, 81 Fulham Road), founded in 2022 by James Drayson (CEO), George Drayson (CAIO), and Sujith Aleshwaram (CTO), with Lord (Paul) Drayson as Chairman. The company builds sovereign AI systems that customers own end-to-end \u2014 model weights, training pipeline, training data, and IP \u2014 rather than renting access through a foreign-controlled API. Its technical core is the patented Forget-Me-Not continual-learning framework, which post-trains open base models (NVIDIA Nemotron, Google Gemma, Qwen) on customer data and then keeps improving them via experience replay. The shipping product family includes the Jupiter series of post-trained LLMs, the Locai One on-premises AI computer (1\u20132\u00d7 NVIDIA RTX PRO 6000 GPUs, from \u00a324,699), the Locai Platform software stack including an OpenAI-compatible API at \u00a36,000/year, the Locai Model Customiser for bespoke model training, an application suite (IDE extensions, chat UI, agentic coding CLI, RAG pipeline, iOS/Android apps), and the GB1 consumer British AI assistant.
The business model combines one-time hardware sales, recurring software subscriptions, bespoke model-training professional services, and managed deployment on Locai Labs\u2019 UK sovereign cloud. Go-to-market is enterprise field sales targeting regulated verticals \u2014 financial services, science and research, energy and utilities, telecoms, manufacturing, public sector, healthcare, legal, security and defence \u2014 where data residency, auditability, and IP ownership are procurement gates. Direct competitors are general-purpose frontier LLM APIs (OpenAI, Anthropic, Google) and hyperscaler offerings; Locai\u2019s substitution argument is sovereignty, fixed cost, and post-trained domain fit rather than raw benchmark superiority. Only one public reference customer is disclosed (First Light Fusion, an Oxford fusion spinout using FLF Coder air-gapped). A flagship April 2026 MOU with Civo under Project Mercury extends the stack to from-scratch pre-trained sovereign foundation models (30B\u2013256B parameters) and is framed around a possible bid into the UK Government\u2019s \u00a3500 million Sovereign AI Fund.
Locai Labs firmographics
Firmographics- Name
- Locai Labs
- Legal name
- Locai Labs Limited
- Website
- https://locailabs.com
- Company type
- Private
- Founded year
- 2022
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Locai Labs is a London-based British foundational AI company that builds sovereign large language models and on-premises AI hardware, post-training open base models on customer data so regulated enterprises and governments can own the model, weights, and IP outright.
- Ownership category
- akta.pro rank
Locai Labs industry classification
Industry- Product category
- Sovereign Enterprise AI Software
- NAICS
- Software Publishers (5132), Custom Computer Programming Services (541511), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Programming Services (7371)
- akta.pro primary industry
- Fine-Tuning, Adaptation & Custom Model Training (PEFT/LoRA/RLHF) (HDAAACAC)
- akta.pro secondary industries
- Model Hosting, Serving & Inference Platforms (HDAAACAB), AI Compute Cloud & GPU-as-a-Service (HDAAAAAK)
Keywords
Where Locai Labs is headquartered
LocationHeadquarters
- HQ city
- London
- HQ country
- United Kingdom
- HQ region
- Europe
Offices2 records
Markets served
Locai Labs business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Infrastructure, Operations, Marketing or Sales, Supply Chain
Revenue model
- Hardware Sales (Locai One): One-time purchase of Locai One hardware units. Base model starts at £24,699 (1 GPU, up to 250 concurrent users), Pro model from £58,000 (2 GPUs, up to 500 concurrent users). Additional storage options at £500 (4TB) and £1,500 (8TB).
- Software Subscription (Locai Platform): Annual subscription for Locai Platform software including API, chat UI, usage platform, and application layer. Priced at £6,000 per year.
- Custom Model Training Services: Bespoke model training services through Locai Model Customiser, where frontier-class models are post-trained on customer proprietary data. Pricing on request.
- Sovereign Cloud Deployment: Deployment on Locai Labs' dedicated UK sovereign cloud for customers who prefer not to use on-premise hardware. Weights, IP and continual-learning artefacts stay with the customer.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| One time/ perpetual license | Pay-as-you-go | Locai One - Entry-level on-prem AI computer |
| One time/ perpetual license | Pay-as-you-go | Locai One Pro - High-performance on-prem AI computer |
| Subscription | Annual | Locai Platform Software |
| Other | Multi-year contract | Bespoke Model (Locai Model Customiser) |
Go-to-market motion2 records
Distribution channels4 records
Marketing channels6 records
Locai Labs product offering
Product offeringCore offering
Locai Labs develops sovereign AI that customers own rather than rent, comprising post-trained large language models (Jupiter series and the 235B-parameter Locai-L1-Large), the Locai One fixed-cost on-premises AI computer, and the Locai Platform software stack (OpenAI-compatible API, chat UI, IDE and mobile applications). The proprietary Forget-Me-Not™ post-training framework enables continual learning on customer data, and bespoke models are delivered via Locai Model Customiser with deployment options on Locai One, the customer's own cloud (AWS/Azure/GCP), or Locai's UK sovereign cloud.
Product overview
Locai Labs offers a platform-plus-modules architecture centered on sovereign AI that customers own rather than rent. The core platform consists of Jupiter (post-trained LLM series), Locai Platform (software stack), and Locai One (on-premises hardware). Locai Model Customiser enables bespoke model training on customer data. An application suite provides IDE extensions (VS Code, JetBrains, Cursor), browser chat interface, agentic coding CLI, RAG pipeline, and mobile apps (iOS/Android). Project Mercury is a pre-trained foundation model series. GB1 is the consumer-facing British AI assistant. Sovereign National AI extends the approach at national scale. All offerings share the core differentiator of customer ownership of model weights, IP, and continual-learning artefacts, with deployment options including on-premises (Locai One), customer cloud (AWS/Azure/GCP), or Locai's UK sovereign cloud.
Differentiator
Problem solved
Functional benefit
Brands
- GB1: The British AI assistant, a consumer expression of Locai's sovereign AI thesis, available at gb1.ai.
- Locai One
- Jupiter
- Project Mercury
- Forget-Me-Not™
- Locai Model Customiser
Products and services
- Locai One Fixed-cost on-premises AI computer that runs sovereign models inside the customer's own environment with no per-token bills. Base unit supports up to 250 concurrent users and models up to 120B parameters; Pro unit supports up to 500 concurrent users and models up to 300B parameters. Targeted at UK and European regulated enterprises, government departments and research organisations that need data residency, auditability and air-gapped operation.
- Locai Platform Annual subscription software stack (OpenAI-compatible REST API, chat UI, control panel, Python SDK) that serves Locai's own sovereign models or any open-source model (Llama, Qwen, DeepSeek, Gemma, Nemotron) to an entire organisation inside the customer's perimeter. API acts as drop-in replacement for api.openai.com/v1 with zero per-token bills and no vendor lock-in.
- Jupiter (Jupiter-N-120B) Sovereign post-trained large language model built on NVIDIA's Nemotron-3-Super-120B base, adapted for British context with Welsh language support, stronger instruction following (+4.1 vs base on IFBench reasoning-on), improved safety (-1.6 on AgentHarm) and +9.1 on Agentic Terminal Bench 2 (medium) versus the Nemotron base. Openly released on Hugging Face and usable inside Locai One or via Locai Platform.
- Locai-L1-Large Flagship British post-trained LLM, a 235B-parameter Mixture-of-Experts instruction model post-trained on a Qwen3 base and aligned with UK laws, languages and cultural context. Achieves +35.4 vs BritLLM on Celtic Languages (BritXNLI 74.48 vs 39.10) and competitive performance against frontier models on human preference (Arena Hard v2 72.90) and instruction following (IFEval 92.45).
- Locai Model Customiser Bespoke model-training service that post-trains a frontier-class base model (Nemotron, Qwen, Gemma or Mercury) on the customer's industry and proprietary data, then deploys the resulting model inside the customer's environment. Customer owns the model, weights and post-training pipeline, eliminating per-token bills and vendor lock-in.
- GB1 British AI assistant positioned as the consumer-facing expression of Locai's sovereign-AI thesis, providing a UK-sovereign chatbot whose outputs reflect the languages of the British Isles and a corpus of Britishness, British information and cultural norms. Available to individual consumers at gb1.ai.
- UK Sovereign Cloud Deployment Managed deployment service on Locai Labs' dedicated UK sovereign cloud for customers who prefer cloud deployment over on-premises hardware. Customer weights, IP and continual-learning artefacts remain with the customer, preserving data residency and IP ownership.
Quantifiable outcome
- Jupiter-N achieves +9.1% improvement on Agentic Terminal Bench vs Nemotron-3-Super-120B base (52.70% vs 43.60%)
- +4 more outcomes
Companies that use Locai Labs
Customer profileNamed customers1 record
Segments11 records
Ideal customer profiles4 records
Locai Labs technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration9 records
AI capability13 records
Feature6 records
Locai Labs partnerships and signals
Strategic signalPartnerships
Five partnerships are on record, tiered flagship and core.
- CivoflagshipLocai Labs and Civo signed a memorandum of understanding to develop the UK's first sovereign large language models through Project Mercury. The partnership will create the Mercury Series of LLMs (30B to 256B parameters) hosted on Civo's UK-based sovereign cloud, providing a local alternative to foreign-owned AI developers. The companies may seek funding from the UK Government's £500 million Sovereign AI Fund. This strategic partnership is core to Locai's sovereign AI infrastructure strategy.
- NVIDIAcoreLocai Labs is a member of the NVIDIA Inception Program. The company post-trains on NVIDIA's Nemotron-3-Super-120B base model for the Jupiter-N series. Locai One hardware uses NVIDIA RTX PRO 6000 GPUs.
- GooglecoreLocai Labs applies its post-training recipe to Google's Gemma-4 base model, creating the Jupiter-G series. Models are openly released on Hugging Face.
- Hugging FacecoreAll Locai models are openly released on Hugging Face including Jupiter-N-120B, Jupiter-G-8B, and Locai-L1-Large, enabling community access and open-source collaboration.
- Qwen (Alibaba)coreLocai-L1-Large is a 235B-parameter Mixture-of-Experts model post-trained on Qwen base, aligned with UK laws, languages, and cultural context.
Scale indicators7 records
Recent moves6 records
Expansion highlights5 records
Locai Labs competitors and assessment
Company assessmentDirect peers
- AI21 Labs: Enterprise foundation model company offering customisable LLMs and on-prem deployment for organisations needing data control. Comparable to Locai on bespoke model training, sovereign/regulated customer focus and developer tooling (SDKs, IDE integrations).
- Mistral AI: Paris-based European sovereign LLM developer offering foundation models, customisation and on-prem/self-hosted deployments. Direct comparison to Locai on sovereign AI positioning and custom post-trained models, though Mistral is significantly larger and better capitalised.
- Aleph Alpha: German sovereign AI company focused on regulated enterprises and public-sector deployments with data residency and explainability. Highly comparable to Locai on sovereign-on-prem positioning and target verticals (government, defence, regulated industries).
- SambaNova Systems: On-prem AI hardware and full-stack platform (DataScale) for running large models inside customer environments with fixed-cost economics. Direct comparison to Locai One on the on-prem AI appliance thesis for sovereign/regulated buyers.
- Together AI: Open-source LLM platform offering fine-tuning, customisation and dedicated/GPU inference for enterprises. Comparable to Locai Platform and Model Customiser on post-training open-source bases and offering on-prem-style deployments.
- Cohere: Enterprise-focused LLM provider offering private deployments, custom models and retrieval/RAG tooling for regulated buyers. Closely comparable to Locai's enterprise field-sales motion, custom model offering and data-residency value proposition.
- H2O.ai: Enterprise AI platform offering on-prem and air-gapped deployments of customisable models for regulated industries. Comparable to Locai on sovereign deployment, vertical focus (financial services, healthcare, public sector) and bespoke model training.
- Nscale: UK-based AI infrastructure provider building sovereign compute and hosting for UK/European AI workloads. Direct comparison to Locai's UK sovereign cloud positioning and Civo partnership for sovereign AI infrastructure in the UK.
Broad incumbents
- Databricks (Mosaic AI): Data and AI platform offering custom model training, RAG and private model serving within customer-governed environments via Mosaic AI. Comparable to Locai on enterprise custom-model and RAG workflow, though part of a much broader data/analytics portfolio.
- Anthropic: Frontier AI lab offering Claude for Enterprise with security/safety focus and increasingly on-prem-style deployment options via cloud partners. A broader incumbent with overlapping regulated-enterprise value proposition though not specialising in sovereign on-prem.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks5 records
Key highlights7 records
Customer concentration
Locai Labs social profiles
Digital presenceLocai Labs compliance and trust
Trust signalCompliance3 records
Locai Labs financial estimates
Financial estimateRevenue estimate
Valuation estimate
Locai Labs leadership team
Management profileNumber of profiles
Profiles4 records
Locai Labs funding detail
Funding detailFunding overview
Funding rounds
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Locai Labs M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Locai Labs
What does Locai Labs do?
Locai Labs develops sovereign AI that customers own rather than rent, comprising post-trained large language models (Jupiter series and the 235B-parameter Locai-L1-Large), the Locai One fixed-cost on-premises AI computer, and the Locai Platform software stack (OpenAI-compatible API, chat UI, IDE and mobile applications). The proprietary Forget-Me-Not™ post-training framework enables continual learning on customer data, and bespoke models are delivered via Locai Model Customiser with deployment options on Locai One, the customer's own cloud (AWS/Azure/GCP), or Locai's UK sovereign cloud.
Is Locai Labs a public or private company?
Locai Labs is a private company. It is classified as founder individual operated bootstrapped and is currently operating.
When was Locai Labs founded?
Locai Labs was founded in 2022. It employs 11 to 50 people.
Where is Locai Labs based?
Locai Labs is headquartered in London, United Kingdom, in the Europe region.
How does Locai Labs make money?
Four revenue lines are on record. Hardware Sales (Locai One) is the primary driver. The others are software Subscription (Locai Platform), custom Model Training Services and sovereign Cloud Deployment.
Who are Locai Labs's main competitors?
Direct peers on record are AI21 Labs, Mistral AI, Aleph Alpha, SambaNova Systems, Together AI, Cohere, H2O.ai and Nscale. Broad incumbents are Databricks (Mosaic AI) and Anthropic.
Does Locai Labs have an API?
Yes. OpenAI-compatible REST API that serves employees frontier-class AI. Drop-in replacement allowing users to switch endpoints from api.openai.com/v1 to their internal infrastructure. Provides access to sovereign models with zero per-token bills and no vendor lock-in. API serves the whole team through Locai Platform with custom inference and concurrency engineering. Developer documentation is at locailabs.com/locai-one.
What industry is Locai Labs in?
Locai Labs's product category is Sovereign Enterprise AI Software. Its primary akta.pro industry code is HDAAACAC, Fine-Tuning, Adaptation & Custom Model Training (PEFT/LoRA/RLHF), with a secondary code of HDAAACAB, Model Hosting, Serving & Inference Platforms. Its NAICS code is 5132 and its SIC code is 7372.