Felix Pinkston
February 13, 2025 11:11
AI uses new serverless APIs and inference clusters to improve DEEPSEEK-R1 deployment to provide high-speed and expandable solutions for large-scale reasoning model applications.
AI introduced significant developments in the distribution of the DEEPSEEK-R1 reasoning model, introducing improved serverless APIs and dedicated reasoning clusters. This move aims to support the increase in demand for companies that integrate sophisticated reasoning models into production applications.
Improved serverless API
The new serverless API of DeepSeek-R1 is known to be twice as fast as other APIs available in the market, allowing for inferences with low speeds and production rates for smooth scalability. This API is designed to provide companies with fast and reactions that have good user experiences and efficient multi -level workflows, which are important for the latest applications that rely on reasoning models.
The main functions of the Serverless API include immediate extensions, flexible payment prices without infrastructure management, and hosted in AI’s data center to improve security. OpenAI compatible API is easily integrated into existing applications, providing high speed limits for up to 9000 requests per minute in the scale layer.
Introduction to reasoning cluster together
In order to compensate for the serverless solution, the AI has started the reasoning cluster together, which provides an optimized GPU infrastructure for the low -through process and intense reasoning. This cluster is particularly suitable for handling a variable and token inferred workloads, achieving up to 110 tokens per second.
The cluster uses an exclusive reasoning engine and is reported to be 2.5 times faster than an open source engine like SGLANG. This efficiency reduces infrastructure costs while maintaining high performance by allowing the same handling with much less GPU.
Expansion and cost efficiency
Together, AI provides a variety of cluster sizes to meet various workloads, and contract -based price models ensure predictable costs. This setting is especially advantageous for companies with mass work rods and offer cost -effective alternatives for token -based prices.
In addition, the dedicated infrastructure guarantees a safe and isolated environment within the North American data center to meet the requirements of personal information and regulations. With its enterprise support and service -level contracts that guarantee 99.9%of operation, AI ensures reliable performance on mission critical applications.
For more information, visit AI together.
Image Source: Shutter Stock