On March 9, the StateSupercomputing Internet PlatformIt was announced yesterday evening that it would go live this weekAlibabaopen source inference model QwQ-32B API interface service, users can getFree 1 Million Tokens.
QwQ-32B is the latest inference model released by Ali Qwen team, built on Qwen2.5-32B + Reinforcement Learning. According to the official benchmark results, QwQ-32B performs as well as DeepSeek-R1 and better than o1-mini and R1 distillation model of the same size in AIME24, which tests mathematical ability, and LiveCodeBench, which evaluates coding ability.
It is reported that the supercomputing Internet completed the QwQ-32B full adaptation in the first time.Includes inference adaptation based on vLLM, PyTorchThe QwQ-32B model image and model files are available on the platform at the same time. Users do not need to download the model locally, and can start the Notebook function with one click to develop the QwQ-32B model on the platform or introduce the proprietary data for further private deployment.
At present, the supercomputing Internet in addition to providing QwQ-32B model API interface services, recently also on-line with DeepSeek-R1 full bucket, the highest 671B full-blooded version of the API deployment and AI Web application services.IT home attached to the national supercomputing Internet platform QwQ-32B API call service steps are as follows:
- Users search for QwQ-32B in the homepage of Supercomputing Internet Mall, click on "QwQ-32B Interface Service", and then click "Go to use" after purchase.
- Enter the API interface information page: Click "Go to use" to enter the API interface information page.
- Choose the form of access on demand, there are three forms of access to choose from: access using http tools such as Postman, Apifox, etc.; access using python code; and direct access on the terminal page.
At present, Supercomputing Internet has connected more than 20 supercomputing and smart computing centers in 14 provinces in China. At present, there are more than 6,500 kinds of arithmetic commodities on the platform, including nearly 240 AI model services, including domestic open-source models such as DeepSeek and Qwen, as well as foreign AI open-source models such as Llama, Stable Diffusion and Gemma.
