The Netherlands
Hardware in the Netherlands. Every model with the glbnxt/ prefix runs here.
GLBNXT Platform / Model Hub
Compare open and frontier models from 14 creators in one catalogue. See where each one is hosted before you send it a single prompt, then call it through one OpenAI-compatible endpoint.
Models from
Data residency, per model
Every model in the hub shows where its hardware sits and who operates it. Filter by region, and keep regulated workloads on Dutch or European infrastructure while teams still get access to frontier models for everything else.
Hardware in the Netherlands. Every model with the glbnxt/ prefix runs here.
Hardware in the EU. The model card tells you who operates it.
US hardware or a US provider. We show that up front, so nobody picks one of these by accident.
glbnxt/ means we run it. Models with this prefix run only on infrastructure GLBNXT operates in the Netherlands.
Find the right model
Search by name or model ID, then stack filters until the shortlist matches your compliance and technical requirements.
Type a model name or its technical ID. Copy the ID to your clipboard in one click.
Name · Model ID · One-click copy
Filter by the organisation behind the model, from frontier labs to open-source research groups.
14 creators
Keep only models hosted where your data is allowed to go.
NL · EU · US
Pick models built for the job, not just the biggest one.
Chat · Coding · Research · Embedding · OCR · Re-ranking · Safeguards
Separate open-weight models you could self-host from closed, API-only models.
Open · Closed
Check the licence before your legal team asks.
MIT · Apache 2.0
Beyond chat
Production AI needs more than a chatbot. The hub covers the full chain, from reading a scanned document to retrieving the right passage and checking the answer.
Assistants, Q&A and drafting for knowledge workers.
glbnxt/gpt-oss-120b · glm-5.3Code generation and agents that work across whole repositories.
glbnxt/Devstral Small 2 · GLM-5.3Reasoning models and web-grounded research with citations.
sonar-deep-research · GLM-5.3Turn documents into vectors for search and RAG pipelines.
glbnxt/qwen3-embedding:0.6bSharpen retrieval by scoring query and passage together.
bge-reranker-v2-m3Read scans, tables and complex layouts into usable text.
deepseek-ocr · mistral-ocr-latestClassify and filter prompts and outputs against your own policy.
gpt-oss-safeguard-20bModel cards
Each model card brings the facts your engineers and your compliance team both need into one view.
✓ Identity and provenance: full name, creator, copyable model ID and the creator's own model card.
✓ Where it runs: regional availability, with redundant regions shown side by side.
✓ Modalities: which inputs it accepts (text, image, audio, video, files) and what it returns.
✓ Supported API parameters, grouped by output control, sampling and tools.
✓ GLBNXT Recommended marks the models we have tested for production use.
One endpoint
The hub sits behind an OpenAI-compatible API. Tools that already speak the OpenAI format connect with a new base URL and a scoped API key.
1 Pick a model in the hub and copy its ID.
2 Create an API key, scoped to only the models that key may use, with an expiry date.
3 Point your code at https://ai.glbnxt.com/v1.
4 Swap models later by changing the model string. Nothing else moves.
Works with n8n · Open WebUI · LibreChat · Langflow · OpenCode · Cline · Aider
Managed by GLBNXT
Open models run on Ollama, managed by GLBNXT inside your EU environment. Weights and requests never pass through an external system.
NVIDIA NIM for latency-critical, high-throughput workloads on NVIDIA GPUs.
Load balancing across instances, automatic failover and priority rules for critical workloads.
Need a model that is not listed, or your own fine-tuned one? We validate, deploy and test it before it appears in your hub.
Models keep improving and your endpoint URL doesn't change. Before a new version takes production traffic, you can run it against your own evaluation set.
Governed by default
The Model Hub connects to the rest of the platform, so every model decision leaves a trail.
The audit trail records which model and version served each request. No silent rerouting to another model.
Token use, time to first token and top models per application and API key.
API keys only reach the models you assign to them, and expire on schedule.
Model identity, provenance and version history support your technical documentation and record keeping.
Which models keep my data in the Netherlands or the EU?
Why are US-hosted models in the hub at all?
Can we add a model that isn't listed, or our own fine-tuned model?
Will a model update break our application?
Do we need to change our code to switch models?
Who decides which teams can use which model?
A 30-minute walkthrough. Bring a workload and we'll shortlist models for it together, Dutch and European options first.