This open-source LLM could redefine AI research, and it’s 100% public

Key Points

1. The open-source LLM by ETH Zurich and EPFL is trained on Switzerland's carbon-neutral 'Alps' supercomputer and will be released under Apache 2.0.

2. The model comes in two configurations: 8 billion and 70 billion parameters, trained on 15 trillion tokens, with coverage in 1,500 languages.

3. The Swiss LLM emphasizes transparency, multilingual scale, and green infrastructure, contrasting with closed 'black-box' systems.

4. The AI market is projected to surpass $500 billion, with over 80% controlled by closed providers, while blockchain-AI is expected to grow from $550 million in 2024 to $4.33 billion by 2034.

5. The Swiss LLM is designed to comply with the EU AI Act, offering advantages in transparency and regulatory compliance, with mandates effective from August 2, 2025.

6. Compared to GPT-4, the Swiss LLM closes the performance gap, especially in transparency and regulatory compliance.

7. Alibaba's Qwen series offers high performance and model diversity, but Switzerland's public LLM focuses on full-stack transparency and multilingual capabilities.

8. Qwen3-Coder has been benchmarked to rival GPT-4 in coding and math tasks, while Switzerland's public LLM leads in multilingual support with 1,500 languages.

9. Switzerland's public LLM runs on a carbon-neutral supercomputer, while Qwen models prioritize speed and scale via Alibaba Cloud.