Contact us
Contact us

AI software and solutions

Aitta ML Inference Service

Through Aitta researchers and developers can run machine learning (ML) model inference on a powerful computing system that is located securely in the EU. When using Aitta for inference, you don’t have to invest in developing and maintaining your own hardware and software solutions for high-performance computing.

Service features

Aitta enables running ML model inference for research and development purposes on the LUMI supercomputer. It provides means for evaluating models’ suitability for various applications with low threshold.

A set of ML models is available including features from image recognition and visual question answering, natural language understanding and generation to predictive analytics through data analysis, identifying patterns and trends. In addition, there are models specialised for embeddings, reasoning, and coding tasks. You can see the models in action and explore their capabilities through your use cases and scenarios. 

Whether your purpose is in research or development, Aitta provides a valuable resource for exploring the latest advances in artificial intelligence and machine learning.

Examples of what Aitta is intended for:

  • testing open source ML models
  • enabling software development projects to develop applications with model inference
  • providing model inference for research projects that need it for experimental purposes

Aitta can be used in several user-friendly ways: with an easy-to-use web user interface, an OpenAI compatible REST API, a Python client-library, or through CSC’s Noppe interface.

The LUMI supercomputer is running in the ISO/IEC 27001 certified datacenter, located in Kajaani, Finland, Europe. 

While Aitta utilizes LUMI resources on demand, the way the supercomputer resources are allocated makes it not suitable for quick initial reaction use cases. Aitta cannot be used for production services to end-users/customers, or processing sensitive or personal data. 

Currently Aitta is offering the following models:

  • google/gemma-4-31b-it, a lightweight, state-of-the-art open model from Google, built from the same research and technology used to create the Gemini models
  • meta-llama/Llama-3.3-70B-Instruct, a multilingual large language model developed by Meta
  • OLMo (allenai/OLMo-7B-0724-Instruct), a series of Open Language Models designed to enable the science of language models
  • openai/gpt-oss-120b, OpenAI’s open-weight model designed for powerful reasoning, agentic tasks, and versatile developer use cases
  • Poro (LumiOpen/Poro-34B-chat), a chatbot version of Poro large language model, a generative pretrained transformer covering Finnish, English and code
  • Poro 2 (LumiOpen/Llama-Poro-2-70B-Instruct), an instruction-following chatbot model created through supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) of the Poro 2 70B Base model
  • TinyLlama/TinyLlama-1.1B-Chat-v1.0, a chat model fine-tuned on top of TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
  • Qwen/Qwen3.6-27B, a powerful coding/vision-language/dense model from Qwen series.
  • Qwen/Qwen3.6-35B-A3B, a powerful coding/vision-language/mixture-of-experts (MoE) model from Qwen series.
  • Qwen/Qwen3-Coder-Next, an open-weight language model designed specifically for coding agents and local development
  • mistralai/Ministral-3-14B-Reasoning-2512, a powerful and efficient language model with vision capabilities, post-trained for reasoning tasks
  • Unbabel/Tower-Plus-9B”, a generative model specialized in translation
  • swiss-ai/Apertus-70B-Instruct-2509, a fully open, massively multilingual and compliant LLM
  • swiss-ai/Apertus-8B-Instruct-2509, a fully open, massively multilingual and compliant LLM

Access/Output Requirements

  • To be able to use Aitta, you will need a LUMI project with a computing package and a user account. 
  • Instructions for creating an account, applying for a LUMI project, and adding members to the project: https://docs.lumi-supercomputer.eu/firststeps/accessLUMI/
  • During the current piloting phase (until the end of September 2026), you can request an invite for using Aitta using the contact form below. Please include in the inquiry message field a description of the aims of using Aitta and what are the expected results or impact of the pilot.
  • Aitta is available for use for as long as the customer has access to LUMI computing resources.

Target groups

Aitta is available for researchers and developers, for public and private sector alike. It can be used for research and development purposes by academic researchers, governmental and other public agencies and commercial businesses.

Note that large businesses cannot use Aitta free of charge through LUMI AI Factory. 

Request process

After a customer has been granted a LUMI computing package, they can start using Aitta.

Pricing

In piloting phase, the use of Aitta is free-of-charge for invited participants.