What's new in Nebius AI Cloud: Q3 2026
Spot pricing for preemptible GPU VMs, new open models and prompt caching on Token Factory, and a web search layer for AI agents. Speakers from Nebius, Token Factory and Tavily walk through the latest updates, with live console demos.
Why watch
If you run GPU workloads, serve open models in production, or build AI agents, Q3 changed what you pay for capacity and how fast you can ship. This session covers the releases in three parts, each presented by a speaker from that product, and ends with a Q&A
What you’ll learn
Nebius AI Cloud platform
- How spot pricing for preemptible VMs works
- How to prototype in DevLabs
- What Cloud Shell adds to the console and how public Nebius Skills give your coding agent platform context
- How Cloud Interconnect, the Wiz integration and audit log export to object storage for your SIEM fit hybrid and enterprise setups
Nebius Token Factory (managed inference)
- Which new open models are live: DeepSeek V4.1 Flash, MiniMax M3.1 Flash (up to 1M context), NVIDIA Nemotron 3.5 Super VL, Reflection’s Beam and Meta’s Muse Spark Open.
- How prompt caching lowers input cost and prefill time for repeated context such as system prompts, tools and code bases
- Where Token Factory ranks on speed and accuracy in independent benchmarks from Artificial Analysis.
Tavily (agentic search)
- How Tavily gives agents live web access, and where it sits in the stack above Token Factory and the cloud platform
- What improved in Q3: query understanding, evidence quality, and index coverage for people and company research
- Which integrations are available
