#Sweden #Amsterdam #Together AI #GPU Infrastructure #Sweden AI

Together AI's European Expansion

Together AI has announced a major step in its European expansion by launching a state-of-the-art GPU infrastructure in Sweden. This new setup is complemented by a series of training workshops aimed at enhancing the skills of AI engineers, with the inaugural Model Shaping workshop set to take place in Amsterdam. These initiatives are pivotal for the company in its pursuit of delivering high-performance, customizable AI solutions to the AI community.

Operational Infrastructure in Sweden

The implementation of the GPU infrastructure in Sweden marks a significant addition to Together AI's global network. This setup fuels the company's serverless inference API, supporting models such as OpenAI’s gpt-oss, DeepSeek, Meta's Llama, and Qwen. Customers in the region can now request dedicated GPU clusters and endpoints directly in Sweden, ensuring secure and low-latency access to services in Northern and Central Europe.

The advantages of the Swedish infrastructure are substantial:
1. Regulatory Compliance: With servers located in Sweden, AI companies can easily meet European data governance and transparency requirements.
2. Reduced Latency: Localizing inference workloads can cut down round-trip network time by 50 to 70 milliseconds for many clients in Northern and Central Europe. This improvement translates to an estimated 25 to 30% enhancement in overall real-time AI application response times.

Mark McKenzie, founder of Caesar (caesar.xyz), highlighted the impact of Together AI’s dedicated endpoints on their operations, stating, "We are using Together's dedicated endpoints with an 8×H200 Llama-4 Maverick deployment to facilitate our transformation phase, as we move towards a public launch. Leveraging the new Swedish region, we expect to decrease latency and meet the data needs of our European clients."

Free AI Development Workshops

In addition to the infrastructure launch, Together AI is excited to introduce a series of free AI development workshops designed for developers. The first event, scheduled for September 10 in Amsterdam, will focus on upgrading and customizing open models. The workshop will touch on key areas like the fine-tuning of supervised training (SFT) on domain-specific data, optimizing preferences, and employing verifiable rewards to achieve performance akin to the leading proprietary large language models at a fraction of the cost.

Additional topics include:

• - Custom Speculative Decoding: To accelerate inference by using optimized draft models for domain-specific pre-generation of tokens, yielding over 1.85× speed-ups on models like DeepSeek-R1.
• - Model Quantization Techniques: To compress large GMLs, minimizing hardware requirements and inference costs, thereby enabling deployments on smaller devices.

As Vipul Ved Prakash, CEO of Together AI, asserts, "Europe is at the forefront of AI innovation, and we are committed to providing builders and researchers with the infrastructure and expertise they need to succeed. Our investments in Sweden and the AI engineer community in Europe embody our dedication to supporting reliable and scalable high-performance AI in the region."

Call to Action

• - Join the Amsterdam workshop for upgrading and customizing open models
• - Request a GPU cluster
• - Apply for a dedicated inference endpoint

About Together AI

Together AI is the leading AI acceleration cloud, enabling developers and AI-native businesses to train, refine, and execute inference for generative AI models—providing unmatched performance, control, and cost-effectiveness. The Together AI platform supports a comprehensive range of the leading open-source and custom models across multiple modalities, while offering flexible deployment options with the highest levels of privacy and security. Committed to advancing the frontier of AI through open collaboration, innovation, and transparency, Together AI ensures that powerful AI systems remain accessible and flexible, producing optimal outcomes for society.