Highlights:

  • Inflection AI has announced its intention to shift focus from its Pi consumer chatbot to creating AI models tailored for the enterprise market.
  • The Inflection for Enterprise appliance combines Gaudi 3 with Inflection AI’s latest large language model, Inflection 3.0.

A well-funded AI startup, Inflection AI Inc., is partnering with Intel Corp. to launch a new appliance for operating large language models.

The companies recently announced their collaboration, which includes an appliance as part of a new offering called Inflection for Enterprise, along with a cloud-based AI service.

Founded in 2022, Inflection AI raised USD 1.3 billion from investors to develop Pi, an alternative to ChatGPT. In March, Microsoft Corp. brought on the company’s Co-founder and CEO, Mustafa Suleyman, to lead its consumer AI group. As part of the deal, Microsoft also hired the majority of Inflection AI’s employees and licensed its AI models in a transaction reportedly valued at USD 650 million.

After the deal with Microsoft, the LLM developer brought in a new leadership team to spearhead a strategic shift. Inflection AI revealed its intention to transition from its Pi consumer chatbot to creating AI models tailored for the enterprise market. This enterprise focus also extends to the new AI appliance that Inflection AI intends to launch through its partnership with Intel.

The appliance is driven by Intel’s Gaudi 3 machine learning accelerator. Launched in April, this processor includes over three times more AI-optimized cores than its predecessor. Intel has also enhanced the integrated Ethernet module that Gaudi 3 uses to communicate with other components within the AI cluster where it’s deployed.

The Inflection for Enterprise appliance integrates the Gaudi 3 with Inflection AI’s latest large language model, Inflection 3.0. According to the AI provider, its software operates on Intel’s hardware with up to twice the cost efficiency of some competing processors.

Inflection 3.0 comes in two versions: one designed for chatbot applications and the other optimized for tasks that require precise adherence to user instructions. The latter version can also format its prompt responses in JSON, simplifying the process for developers to integrate the model’s outputs into their applications.

Customers of Inflection for Enterprise will receive a version of Inflection 3.0 tailored to their specific needs. The company customizes the LLM by fine-tuning it with each organization’s data. According to Inflection AI, this fine-tuning process enhances the model’s utility for employees and ensures that its output adheres to the organization’s internal content style guidelines.

“Inflection for Enterprise is the only system that allows enterprises to own their intelligence in its entirety. You own your data, your fine-tuned model, and the architecture it runs on. It’s fully in your control to host on-premises, in the cloud, or hybrid,” Sean White, Inflection AI CEO, wrote in a blog post.

Intel and Inflection AI aim to launch their jointly developed appliance in the first quarter of 2025, with the chipmaker expected to be one of the initial customers. In the meantime, Inflection for Enterprise is accessible through Intel Tiber AI Cloud, a platform that offers on-demand access to Gaudi 3 and several other processors from Intel.