# Highlights - May 2026

• Introduced à-la-carte IaaS pricing: customers can now purchase compute (CPU), memory (RAM), storage, and GPU resources individually, per unit, per month.
• Existing customers stay on current plans during the transition period; resource additions continue via support tickets until migration is completed.
• Updated documentation for the new IaaS offer: how the service works, getting started, platform architecture (OpenShift capacity, namespaces), and resource pricing.
• Model service: active models table and guidance updates (context limits, types, use cases).
• Updated Model Service subscription documentation for the Evaluation and Exploratory plans.
• Updated PHOENIQS Chat documentation: Quick Start, Build your own workflows, and subscriptions.
• Decommissioned MaaS models scheduled for retirement on 28.05.2026: inference-apertus-8b, inference-deepseekr1-70b, inference-deepseekr1-670b, inference-kimi-k2, inference-llama33-70b, and inference-qwq25-vl-72b; moved to the decommissioned models list.
• Removed inference-kimi-k26 (Kimi K2.6) from sandbox evaluation after it did not pass testing; the model requires NVIDIA B200/300-class GPUs, which are not yet available in our infrastructure.
• Added inference-GLM-51 (GLM 5.1) to sandbox models for performance, stability, and cost validation; promotion to the active models catalog is pending successful evaluation.