Blog
Engineering notes on running open-weight models in production: capacity pricing, routing, retrieval, and the layers around inference.
Unlimited Basic upgrade: Qwen 3.8 27B, GLM and DeepSeek
Unlimited Qwen 3.8 27B and 5,000 monthly AI Credits for GLM-5.3-Flash and DeepSeek for €14.95. Explore model benefits, official benchmarks and API access.
QDivZero Unlimited Basic: Qwen, GLM and DeepSeek
Unlimited Basic for €14.95/month: unlimited Qwen 3.8 27B subject to fair use plus 5,000 shared monthly AI Credits for GLM and DeepSeek.
Sherry 35B 0.1: controllable reasoning for code, analysis, and hard tasks
Introducing Sherry 35B-A3B 0.1, a GRPO preview with low, medium, and high reasoning effort, measured improvements on code and knowledge tasks, and a transparent self-distillation training recipe.
QDivZero is now in public beta
The story of QDivZero from its private beta to a public AI infrastructure platform for teams that want more choice and control.
Valendra joins NVIDIA Inception to advance QDivZero
Valendra joins NVIDIA Inception with QDivZero, the platform for deploying open-weight AI models without managing the underlying GPU infrastructure.
Qwen 3.8 27B will be available to everyone for one week
Qwen 3.8 27B is available to everyone on QDivZero for seven days as we continue learning from real-world inference workloads during our private beta.
QDivZero Storage brings persistent disks to Compute
Storage extends Compute with persistent disks that can be mounted into multiple workloads, shared across launches, and kept attached when a workload moves between providers.
random-moe-pruning: dataset-free expert pruning for Mixture-of-Experts checkpoints
A Valendra research artefact for shrinking Mixture-of-Experts checkpoints through expert pruning, without a calibration dataset and reproducible on CPU. Current release validated on GPT-OSS.
QDivZero opens private beta for capacity-priced AI infrastructure
The private beta for QDivZero is now open: one operating layer for model execution, routing, retrieval, and guardrails without per-token billing.