The Linux Foundation is taking charge of the Open Model Initiative (OMI). This initiative aims to develop free and open-source LLMs that remain accessible to all. In addition, the collective wants to focus on clear ethical and shared standards.
With the Open Model Initiative, the Linux Foundation wants to take a clear step in providing truly “open” GenAI models and thereby respond to the growing demand for them.
OMI should respond to the retreats that some AI players have made, such as OpenAI. Other companies have also curtailed the deployment of GenAI by adjusting licensing requirements.
For example, the promoters point to Stability AI’s recent modifications for its popular image generation model Stable Diffusion 3 (SD3). This LLM was previously available indefinitely, but now requires a monthly subscription and has other usage restrictions. Also, Stability AI is not clear on licensing terms.
OMI was founded by a number of organizations, including Invoke AI, CivitAI and Comfy Org, as well as receiving support from community projects such as Sentient Foundation and Wand Synthesis AI. Within the Linux Foundation, the open source AI initiative has now received the status of an official “open source foundation.
Removing barriers
The openness focuses primarily on training LLMs and developing them with irrevocable open licenses. They are intended to always remain without removal clauses and access fees.
For this, OMI first wants to establish a governance framework and specific working groups to encourage joint development. The Linux Foundation also wants to conduct a survey to get feedback from the AI community on future model research and training.
In addition, the AI initiative wants to establish shared standards to improve interoperability between LLMs and metadata operations. A transparent open-source dataset for AI training purposes that all developers can use is additionally on the horizon.
Furthermore, OMI plans to develop an alpha test model for targeted red-teaming. A first version of this model, along with scripts to fine-tune the model, should be available before the end of this year.
OPEA initiative as well
OMI is not the first open source AI initiative the Linux Foundation has supported. Earlier this year, the open source organization already took the Open Platform for Enterprise AI (OPEA) under its wing. Among other things, this initiative develops open standards for Retrieval-Augmented Generation (RAG).
Anyone can contribute to OMI. Among other things, the initiative has prepared invitations to do so in its official GitHub repository and through their Discord channel.
Also read: OpenELA automates enterprise Linux source delivery