# Highlights - October 2026

Added S3 Object Storage, a new IaaS offering for S3-compatible object storage, covering portal purchase, ObjectBucketClaim provisioning on OpenShift, and connecting clients.

Removed inference-minimax-h3 (MiniMax H3) from Preview Models.

Set the Preview End Date for inference-kimi-k3 (Kimi K3) to 16.11.2026 on Preview Models.

Set the Disposition of inference-kimi-k3 (Kimi K3) to "Not Planned for Hosting" on Preview Models.

Corrected the Context window of inference-kimi-k3 (Kimi K3) to 512K on Preview Models.

Corrected Kimi K3's Model Name on Preview Models to inference-kimi-k3-preview, the identifier actually available on the Model Service API.

Updated inference-glm5 (GLM 5.3) on Preview Models: Preview End Date 16.11.2026, disposition Production Hosting, pricing 1.1667 / 3.6667 Credits/M Tokens (input / output), context window 131K.

Added inference-deepseek-v4-flash (DeepSeek V4 FLASH, Version 4.1) to Preview Models for performance, stability, and cost validation; Preview End Date is 16.11.2026 and the disposition is Production Hosting. Hugging Face reference: deepseek-ai/DeepSeek-V4.1-Flash-1017.

Corrected GLM 5.3's Model Name on Preview Models to inference-glm5-preview, the identifier actually available on the Model Service API.

Updated the Model Name of DeepSeek (V4.1 Flash) on Preview Models to inference-deepseek-v4-flash-preview.

Added inference-glm5-flash-compact (GLM 5.3 Flash Compact) to Active Models on the Phoeniqs Model Service. Same model as GLM 5.3 Flash with a smaller (512K) context window. Hugging Face reference: zai-org/GLM-5.3-Flash.