Aleph Alpha

Overview

Aleph Alpha is a European provider of large language and multimodal models, developed with a focus on enterprise requirements, including data sovereignty, explainability, and compliance with regulations such as GDPR. Founded in 2019, the company offers a suite of models under the Luminous brand, designed to process and generate content across various modalities, including text, images, and speech. This multimodal capability distinguishes their offerings, allowing for more complex AI applications that integrate different types of data inputs and outputs.

The Luminous model family includes variants like Luminous Base, Extended, Supreme, and World, each offering different scales and capabilities to suit diverse enterprise needs. These models are accessible via an API, with SDKs available for Python and JavaScript, facilitating integration into existing systems. Aleph Alpha emphasizes the ability for customers to maintain control over their data, offering deployment options that range from cloud-based access to on-premise or private cloud installations, which can be critical for organizations with strict data governance policies.

Aleph Alpha positions its technology for use cases requiring robust, auditable, and transparent AI solutions. This includes applications in sectors like finance, healthcare, and government, where explainability and data security are paramount. The platform provides tools and features aimed at enhancing the interpretability of model decisions, which can assist developers and technical buyers in understanding and validating AI outputs. The company's commitment to these principles aligns with the increasing demand for responsible AI development and deployment, particularly within the European regulatory landscape, as outlined by organizations like the European Commission in their AI Act proposals.

For developers, the platform offers a developer experience noted for its clear API references and Python SDK examples. This is intended to simplify the process of integrating Aleph Alpha's models for common LLM tasks, as well as leveraging their multimodal functionalities. The pricing model is pay-as-you-go based on token usage, with provisions for volume discounts and custom enterprise arrangements, providing flexibility for organizations scaling their AI initiatives.

Key features

• Multimodal AI Models: Processes and generates content across text, image, and speech modalities, enabling applications that integrate diverse data types.
• Luminous Model Family: A suite of large language models (Luminous Base, Extended, Supreme, World) offering varying capacities for different enterprise use cases.
• Data Sovereignty Options: Provides deployment flexibility, including cloud, on-premise, or private cloud, to help organizations maintain control over their data.
• Explainable AI (XAI): Features designed to increase the interpretability and transparency of model decisions, assisting in understanding and validating AI outputs.
• Enterprise-Grade Security: Built with a focus on security and compliance, including adherence to GDPR standards for data protection.
• API Access and SDKs: Offers programmatic access via a REST API, supported by official SDKs for Python and JavaScript, to facilitate integration.
• Fine-tuning Capabilities: Allows for adaptation of general models to specific domain data and tasks, enhancing performance for specialized applications.

Pricing

Aleph Alpha's pricing model is pay-as-you-go, primarily based on token usage. The company offers volume discounts for higher usage tiers and provides custom enterprise pricing for large-scale deployments or specific requirements. A free API access tier with limited credits is available for developers to explore the platform. As of May 2026, the general pricing structure is as follows:

For detailed and up-to-date pricing information, refer to the official Aleph Alpha pricing page.

Common integrations

• Python Applications: Integrate Luminous models into Python-based backends, data science workflows, or web applications using the Aleph Alpha Python SDK.
• JavaScript/TypeScript Frontends and Backends: Utilize the Aleph Alpha JavaScript SDK for client-side applications or Node.js services requiring AI capabilities.
• Cloud Platforms: Deploy and manage Aleph Alpha models within various cloud environments, including private cloud setups, to align with data sovereignty requirements.
• Enterprise Data Lakes & Warehouses: Connect to existing data infrastructure for fine-tuning models or processing large datasets with AI.
• Custom Business Applications: Embed AI functionalities into bespoke enterprise software for tasks like content generation, semantic search, or intelligent automation.

Alternatives

• OpenAI: Offers a range of large language models like GPT series, known for broad general-purpose AI capabilities.
• Anthropic: Focuses on developing safe and steerable AI systems, including their Claude family of models.
• Google Cloud AI: Provides a comprehensive suite of AI and machine learning services, including Vertex AI for model development and deployment.

Getting started

To begin using Aleph Alpha's Luminous models, you typically start by obtaining an API key and then using one of their SDKs. The following Python example demonstrates how to make a basic text completion request.

import os
from aleph_alpha_client import Client, Prompt, CompletionRequest, TokenizationRequest, ExplanationRequest, Text, Image, MultimodalPrompt

# Set your Aleph Alpha API token
# It's recommended to load this from an environment variable for security
# os.environ["AA_TOKEN"] = "YOUR_ALEPH_ALPHA_API_TOKEN"

try:
    token = os.environ["AA_TOKEN"]
except KeyError:
    print("Please set the AA_TOKEN environment variable.")
    exit()

client = Client(token=token)

# Example 1: Text Completion
print("\n--- Text Completion Example ---")
prompt_text = "The capital of France is"
request_completion = CompletionRequest(
    prompt=[Text(text=prompt_text)],
    model="luminous-base", # Or luminous-extended, luminous-supreme, etc.
    maximum_tokens=10,
    temperature=0.7,
    top_k=0,
    top_p=0,
    repetition_penalties=False,
    tokens_expected=10
)

response_completion = client.complete(request_completion)
print(f"Prompt: '{prompt_text}'")
print(f"Completion: '{response_completion.completions[0].completion}'")

# Example 2: Multimodal Prompt (Text and Image - conceptual, requires actual image data)
# This is a conceptual example. For a real use case, 'path_to_image.jpg' would be a local file path
# and the image would be loaded using Image.from_file().
# print("\n--- Multimodal Completion Example (Conceptual) ---")
# multimodal_prompt = MultimodalPrompt(
#     prompt=[
#         Image.from_file("path_to_image.jpg"),
#         Text(text="Describe the image in detail:")
#     ]
# )
# request_multimodal_completion = CompletionRequest(
# #    prompt=multimodal_prompt,
#     model="luminous-extended",
#     maximum_tokens=50,
#     temperature=0.7
# )
# # response_multimodal = client.complete(request_multimodal_completion)
# # print(f"Multimodal Completion: {response_multimodal.completions[0].completion}")
# print("Multimodal example requires a valid image file and deeper setup.")

# Example 3: Tokenization
print("\n--- Tokenization Example ---")
text_to_tokenize = "Hello, world! This is a test."
request_tokenization = TokenizationRequest(
    prompt=[Text(text=text_to_tokenize)],
    model="luminous-base",
    compress_to_gb_tokens=False,
)
response_tokenization = client.tokenize(request_tokenization)
print(f"Original text: '{text_to_tokenize}'")
print(f"Tokens: {response_tokenization.tokens}")
print(f"Number of tokens: {len(response_tokenization.tokens)}")

This code snippet demonstrates how to initialize the client with your API token and perform a simple text completion. For multimodal capabilities, you would extend the Prompt to include Image or other modalities as described in the Aleph Alpha API reference. Ensure your API token is securely managed, ideally through environment variables, as shown in the commented section.