Intel has released a public beta of its enterprise hybrid AI platform SuperClo based on the open-source OpenClo framework, HotHardware reported on July 9 local time.
SuperClo is based on an architecture that dynamically distributes work between local systems and the cloud.
High-frequency tasks such as local file parsing, memory search and masking sensitive data are handled entirely on-premises. Only complex multi-step reasoning or deep web searches are selectively sent to cloud large language models. Intel said local models process more than 70 percent of tokens, sharply reducing fees for Claude or ChatGPT.
In internal corporate workload tests cited by Intel, SuperClo hybrid routing can cut average cloud token consumption by more than 70 percent. If so, it could lower the cost barrier that many companies have cited as a reason for scaling back AI rollouts and change the economics of companywide AI adoption, HotHardware said.
System requirements for the public beta are fairly high. Intel said it requires a Panther Lake laptop with 16 gigabytes of RAM and a workstation system equipped with four Arc Pro B70 GPUs.
Chi-gyu Hwang
delight@d-today.co.kr