Nous Research Introduces an API for Unique AI Models Not Offered by OpenAI or Anthropic
Subscribe to our daily and weekly newsletters for updates and exclusive content on industry-leading AI developments. Learn More
Nous Research Launches Inference API
Nous Research, an AI collective based in New York, has unveiled a new Inference API designed to enhance accessibility for developers and researchers. This programmatic interface marks a significant enhancement to the company’s services and positions it as a competitor to larger AI firms like OpenAI and Anthropic.
Following community feedback, Nous Research has constructed a user-friendly system aimed at making its language models more available to a broader audience. This API launch features two significant models: the Hermes 3 Llama 70B, a general-purpose model built on Meta’s Llama 3.1 architecture, and the DeepHermes-3 8B Preview, which enables users to switch between typical responses and in-depth reasoning.
Managing Demand with a Waitlist System
To cope with the anticipated demand for the API, Nous Research has introduced a waitlist approach at their portal portal.nousresearch.com. Access is granted on a first-come, first-served basis, and every new account receives $5.00 in complimentary credits. Developers can consult the API documentation for integration details.
This waitlist strategy reflects Nous Research’s unique position in the AI landscape. Unlike large firms with vast GPU resources, Nous faces the typical limitations of smaller entities in AI, necessitating a careful management of their computational infrastructure. This strategic waitlist serves not only as a practical solution but also acts as a marketing tool, creating an air of exclusivity that generates interest.
Transitioning from Open Source to a Cloud API
This new API comes just four months after the launch of Nous Chat, the organization’s first user-facing chatbot. Previously, developers had to download and run the models locally, which was often a complicated and costly process. The API, however, allows access to high-performance versions of these models without requiring developers to manage their own infrastructures.
DeepHermes-3, introduced last month, exemplifies the company’s push into the competitive realm of reasoning-focused AI, allowing users to alternate between quick answers and detailed thought processes via a prompt. This reflects the company’s growth strategy while addressing market demands.
Unrestricted AI and its Philosophy
Since its establishment in 2023, Nous Research has advocated for an alternative to the more controlled AI ecosystems offered by larger companies. They push for individual agency and alignment with user preferences, as seen in their blog posts, such as “Freedom at the Frontier” and “From Black Box to Glass House.”
The philosophy underpinning Nous Research emphasizes freedom and individual empowerment, asserting that the advancement of superintelligence should not be confined to a select few corporations. While the company showcases a commitment to unrestricted AI, it has also integrated some safeguards to prevent harmful outputs, balancing its ideals with responsibility.
A New Business Model for AI
The launch of the Inference API reflects Nous’ transition towards a sustainable business model while still adhering to open-source principles. Since July 2023, the company has released 29 different AI resources, including models, datasets, and research papers.
This new API indicates a significant evolution in Nous Research’s approach, as they explore commercialization without compromising the open-source community that supports them. Their strategy is designed to cater to diverse market demands: individual developers can still download models, while businesses can opt for reliable API access.
Nous Research aims to expand its API offerings over time, potentially incorporating additional models like Hermes 2 Pro and its new Psyche project. This approach may provide a fresh alternative in an increasingly competitive environment, fostering innovation in AI inference.
The company remains open to community feedback, inviting collaboration to shape the future of AI development.
