Robots Atlas>ROBOTS ATLAS
Luminous

Luminous

Luminous · Family: Luminous
Aleph Alpha’s LLM family (13B–70B; base/extended/supreme): multilingual, multimodal (text+image), with explainability (AtMan) and “sovereign AI”.
⚠ Deprecated🏢 EnterpriseLLMMultimodal📁 Luminous
Context window
2K–8K (base 2K, extended 8K)
tokens
Parameters
13B–70B (base 13B, extended 30B, supreme 70B)
parameters
Release date
12 October 2022
Access:APIDeployment:☁ Cloud💻 Local

Overview

Luminous is a family of large language models (LLMs) developed by the German company Aleph Alpha, headquartered in Heidelberg, released on 12 October 2022. The models are built on a transformer (GPT) architecture with self-supervised learning, aimed at enterprise and public-administration use (“sovereign AI”).

The family includes variants differing in scale: Luminous-base (~13B parameters, 2K context window), Luminous-extended (~30B, 8K window), Luminous-supreme (~70B) and the instruction-tuned Luminous-supreme-control. The base and extended variants are multimodal — accepting text and images as input.

A distinctive feature of Luminous is explainability: Aleph Alpha provides methods (including AtMan) that make the model’s decisions more understandable, addressing the “black box” problem. The models are multilingual (including English, German, French, Spanish and Italian) and were offered via an API and in local/on-premises deployments compliant with European data-protection standards.

Luminous is a proprietary model aimed at enterprise and public-sector customers. Aleph Alpha (founded in 2019 by Jonas Andrulis and Samuel Weinbach) has been backed by Schwarz Gruppe, Bosch and SAP, among others. The Luminous family was subsequently succeeded by the Pharia models.

Classification
LLMMultimodal
Family: Luminous
Access & deployment
API
CloudLocal
Weights: Closed
Key parameters
📏 Context: 2K–8K (base 2K, extended 8K)
🧩 Parameters: 13B–70B (base 13B, extended 30B, supreme 70B)
✓ Fine-tuning
📥 Input: text, image

Technical specification

Context window
2K–8K (base 2K, extended 8K)
tokens
Parameters
13B–70B (base 13B, extended 30B, supreme 70B)
parameters
License
Proprietary
Hardware requirements
Available via the Aleph Alpha API and in local/on-premises deployments (sovereign AI).
Features:Fine-tuning
Modalities
⬇ Input
textimage
⬆ Output
text

Capabilities and applications

Native model capabilities
Language modeling
Ability to predict subsequent tokens and generate coherent natural-language text based on the preceding context.
Category: language
Multilingual
Competence in many natural languages (from a few to over a hundred): understanding, generation, translation, and code-switching within a single conversation. Frontier models support a wide range of languages with comparable quality.
Category: language
Multimodal understanding
Category: multimodal
Image understanding
Analysing and interpreting the content of images.
Category: vision
Interpretability
Explainability of model predictions: identifying features driving the decision (SHAP, feature importance, attention weights, counterfactuals).
Category: other
Instruction following
Precisely following instructions contained in the prompt: response format, length, style, constraints (e.g. 'reply in six words'). GPT-5.1 significantly improved this capability compared to GPT-5.
Category: language

Technical architecture

Core Architecture

Deployment and security

🔒 Security / Enterprise
✓ Verified enterprise information
Updated: 24 Jul 2026↗ Security documentation