Russian cloud provider Turbo Oblako, part of the RTK-TsOD Group, has updated its services for developing and running artificial intelligence solutions. DeepSeek-V4.1-Flash and GLM-5.3 language models have been added to the Foundation Model Hub catalog, model scheduling configuration has appeared in Inference Platform, and the GPU lineup has been expanded with the NVIDIA RTX PRO 6000 Blackwell Server Edition accelerator.
DeepSeek-V4.1-Flash processes text and images, supporting a context window of up to 1 million tokens. GLM-5.3 is designed to work with text and features a context window of up to 400,000 tokens. Both models are available via API with pay-per-token pricing, without the need to deploy infrastructure independently.
In Inference Platform, users can set availability periods for their custom models taking time zones into account. Once the specified interval ends, the GPUs are shut down, and no fee is charged for their use. This helps reduce computing resource costs when AI services are not in demand.
The new NVIDIA RTX PRO 6000 Blackwell accelerator is equipped with 96 GB of GDDR7 VRAM with ECC error correction support and a bandwidth of 1.6 TB/s. Up to eight such GPUs can be deployed in a single virtual machine. The accelerator is designed for running large language models, fine-tuning them, and handling other compute tasks. It complements the hardware lineup, which also includes NVIDIA L4, L40S, and H200.