Recently, the large language model service platform, Volcano Engine's "Volcano Ark", officially announced its integration with the latest version of DeepSeek-R1-0528. This move not only highlights Volcano Engine's technical strength in the field of large language model services but also provides enterprise users and developers with a more efficient and convenient experience in applying large language models.
The Volcano Ark platform has built a high-performance service system targeting the core needs of large language model applications: speed and stability. By using the self-developed xLLM high-performance inference framework, the platform achieves an extreme inference speed of 30 milliseconds per token, and ensures industry-leading stability, maintaining high-efficiency low-latency output even under load fluctuations to ensure smooth real-time interactions. In addition, Volcano Ark provides massive concurrent support by default, supporting 5 million TPM (Tokens Per Minute) and 30 thousand RPM (Requests Per Minute), fully meeting the demands of enterprise-level high-concurrent calls and effectively avoiding service interruptions during traffic peaks.
In terms of scenario coverage, the Volcano Ark platform offers practical function supports such as Function Call and internet connectivity for the DeepSeek-R1-0528 model. It builds a comprehensive support system for the diverse scene requirements of enterprises and developers in actual applications. Whether it is offline batch inference to handle large-scale data processing scenarios or prefix caching technology to improve response speeds in applications involving repetitive prompts or standardized beginnings, Volcano Ark can provide flexible and efficient solutions. Additionally, the platform offers a TPM assurance package that allows users to dynamically adjust traffic quotas based on business peak demand, ensuring uninterrupted service in critical scenarios.
To help enterprise users and developers quickly get started and smoothly implement large language model applications, Volcano Ark provides multiple access points for diverse experiences. In the Volcano Engine Experience Center, users can directly experience the core functions of the new DeepSeek-R1-0528 model for free without registration or login. After verifying the effects, they can seamlessly transition from "tasting experience" to "official call" by one-click jump to the console for registration and configuration. For professional developers, the official Volcano Ark console provides efficient configuration tools such as quick model call configurations, API direct connections, and visual parameter debugging. The Application Lab also opens-source high-value large language model application templates covering basic to complex scenario needs, providing enterprises with ready-to-use toolkits.
Notably, Volcano Ark also launched a 50% discount promotion for new customers, offering an ultra-low price to assist new users in quickly starting their exploration of large language models. Enterprises or individuals who have not registered for a Volcano Engine account can enjoy a 50% discount on the usage of up to 1 billion tokens of the DeepSeek-R1-0528 model within 14 days after registering and completing identity verification, while keeping the original price unchanged.