TL;DR. As of today, azena is an official Nebius partner. Our Model Garden has been running on Nebius Token Factory since early October. For our customers this means one access point for open models and for Claude, with every request assigned to a customer and billed against a cap.
What Nebius is
Nebius builds a cloud made for AI, in its own words “engineered from silicon to API”. With Token Factory, Nebius serves open language and image models through a simple API that scales with the load. That is what we build on.
What runs on it at azena
The Model Garden is our own access point to AI models. One key covers open models such as Kimi K3 as well as Claude. Costs are kept separate per customer, and every customer has their own cap. Our Brand Studio has also been building its presentations through the Model Garden since October 10.
Why open models
Because they hold their own on real work. We had models build five real presentation briefs and compared the results blind, judging both orders each time.
Blind pairwise comparison · 5 presentation briefs · 10 verdicts each
Kimi K3 on Nebius won all ten verdicts against GPT-6 Sol. Only Claude Opus 5.5 beat Kimi K3, and narrowly at 6 to 4. That is why we use open models where they keep up, and the expensive ones where they are measurably better.
What it means for you
- One access point, many models. You pick the right model for each task without managing five contracts and keys.
- Costs under control. Every request belongs to a customer or project, with a fixed monthly cap.
- Measured, not assumed. Which model takes on a task is decided by blind comparisons on real briefs.
How we measure models is covered in detail in Model Garden: how we measure AI models. Thanks to the Nebius team. If you want to use open models in your own business, we are happy to show you our setup.
