10 million users.
100+ models tested every week.
More than 1 trillion tokens processed every day.
Henry Wang explains why he thinks the next generation of consumer AI won't be won by the best model, but rather by the teams that can iterate and scale the fastest.
Listen now:
Nebius is at ICML 2026 in Seoul (July 6–11), booth B713. Meet the team behind two papers: SWE-rebench V2, a 36,000+ task open dataset for coding agents, and LK Losses, faster LLM inference. We're hiring too... nebius.com/events/icml-20…
We’re back with more data center ASMR. This one's for anyone whose happy sound is the slide and click of a server seating into a rack. Fresh tracks every Friday through July.
Kimi-K2.7-Code is live on Nebius Token Factory.
@Kimi_Moonshot 's latest coding-focused open-weight model is built for long-horizon software engineering: repo-scale changes, multi-file refactors, debugging loops, and tool-heavy coding agents.
A strong fit for teams testing
Scientific AI & healthcare is a priority for Nebius. The AI Discovery Awards is one way we're proud to invest in the ecosystem.
See all 20 winners: nebius.com/newsroom/nebiu…
The Nebius AI Discovery Awards 2026 wrapped in London last night. 647 applications, up from 257 last year. 20 winners across scientific and healthcare AI.
28 judges from NVIDIA, AstraZeneca, AbbVie, Stanford, Accel, Index Ventures, Sofinnova, Cathay Innovation, the NHS, and
At @PrimaMente, speed matters. Every failed experiment, delayed training, or lost week delays delivering Alzheimers prevention, treatment, or cure to patients.
That’s why they turned to Nebius to speed up training cycles for their foundation models.
Full story →
How do you spot an AI unicorn before it has any revenue?
@brianzhan1 of @strikervp has a framework. And it doesn't involve business plans.
Hear it on the Nebius for Startups Podcast → nebius.com/podcast
We made ASMR videos out of real data center sounds. If you find the click of a memory stick perfectly seating in a motherboard unreasonably calming, today is your lucky day. More tracks dropping every Friday through July.
youtu.be/KI55D-S9Yas
What’s your weirdly calming tech
Today we’re launching Custom Speculator Training in Nebius Token Factory.
Train workload-specific draft models from your own data and deploy them alongside the base model.
Generic drafters are a good start, but production traffic is rarely generic 👇
The Nebius Certification Program is live.
Whether you're the engineer keeping GPU clusters running, the architect designing the infrastructure, or the team lead who wants to know your stack is in good hands, there's a track for you.
Learn more at: nebius.com/certification
Nebius is coming to Stockholm on July 1, with Anyscale. If you’re an AI/ML engineer, come compare notes on today’s biggest challenges: idle GPUs, NCCL debugging, inference that collapses under load.
Deep-dives, drinks, discussion, and debugging.
📍 Convendum · Jul 1 · 17:00 →
Introducing Aether 3.6, the latest version of the Nebius AI Cloud platform.
Every feature in it came from our customers’ evolving needs: workload security, cost and control improvements, developers’ day-to-day operational simplicity.
nebius.com/blog/posts/ai-…
