Common questions on Apertus
What makes Apertus different from other open models?
Apertus is fully open — offering complete access to training data, code, and alignment principles, not just open weights. This openness supports research, collaboration and informed use, creating a valuable tool for the global AI community. See Swiss AI Initiative.
Is Apertus compliant with the EU AI Act?
Designed for compliance, Apertus is built to meet EU AI Act requirements from day one, including opt-out mechanisms, PII removal, and memorization prevention. Day-to-day compliance is still the responsibility of users deploying Apertus. Read Tech Report section 3.1 and Fan et al., 2025 for details.
May I use Apertus commercially?
Yes. Apertus is released under Apache 2.0, allowing both research and commercial use without restrictions or royalties. Use it freely in any project, but abide by license terms and be aware of statutory limitations. Try the Open License Helper.
What languages does Apertus support?
Apertus is trained on 1000+ languages, making it a strong foundation for further development. Currently, Apertus is conversational in a few dozen languages. For specific language needs, we encourage users to evaluate or fine-tune Apertus. Refer to our language support documentation for more details.
Can I fine-tune Apertus?
Yes! You can fine-tune Apertus with your own data for specialized tasks (e.g., domain-specific knowledge, proprietary data). Guidance and tools are available in our fine-tuning guide.
How is Apertus aligned?
Apertus was trained with ethical and safety principles to align with Swiss values and EU AI Act goals (e.g., transparency, fairness, human oversight). While we strive to minimize bias, users should be aware of inherent biases in language data and test for fairness in their applications. See the Swiss AI Charter for details.