Berget AI Launches Berget Code: Swedish-Hosted AI Coding Assistant on Open Models
Key Takeaways
- ▸Berget Code becomes the first European coding assistant to use Kimi K2.6, proving open models can compete with proprietary American tools at production quality
- ▸Swedish hosting and GDPR compliance eliminate CLOUD Act exposure, making it compliant with European regulatory requirements and data sovereignty mandates
- ▸OpenAI API compatibility enables frictionless migration for developers currently using Claude Code or other American services without retraining
Summary
Berget AI, a Stockholm-based sovereign AI infrastructure company, has launched Berget Code, a new coding assistant service designed for European developers who prioritize data security and regulatory compliance. The service runs on open-source software and open models, with Kimi K2.6 from Moonshot AI as the flagship model, offering a 256,000-token context window that enables analysis of entire codebases in a single session.
Berget Code positions itself as a direct alternative to American coding tools like Claude Code and Codex, addressing European developer frustrations with rising prices, availability constraints, and sustainability concerns from U.S. providers. The service is fully hosted on Swedish infrastructure, ensuring all data remains within Sweden's borders, with GDPR compliance and immunity from the U.S. CLOUD Act—a critical differentiator for government agencies, financial institutions, and security-conscious organizations.
The launch is significant timing: open language models have recently achieved parity with proprietary services for coding tasks, removing the previous quality trade-off of choosing sovereignty over capability. Berget Code supports multiple open models (Kimi K2.6, Gemma 4, Mistral Medium 3.5), maintains OpenAI API compatibility for seamless migration, and commits to not training its models on user code.
- 256,000-token context window allows analysis of entire codebases, matching or exceeding capabilities of leading American coding assistants
- Launch reflects broader shift toward regional AI infrastructure as open models reach feature parity with closed systems
Editorial Opinion
Berget Code represents a meaningful inflection point in the AI infrastructure market: open models have finally reached production-grade quality for coding tasks, making them viable alternatives to proprietary American services. For European organizations, this launch addresses a real regulatory and geopolitical pressure point—the tension between adopting cutting-edge AI tools and maintaining data sovereignty. Berget AI's OpenAI API compatibility is particularly smart, lowering switching costs and making the business case more compelling. This signals the beginning of regional AI consolidation, where regulatory fragmentation and data protection concerns drive demand for locally-hosted alternatives.
