Why small models are making a big impact in AI accessibility

Building AI together: Three open source projects creating a more transparent future.

Sriram Raghavan, vice president of AI research at IBM, sat down with Mark Hinkle, co-founder All Things AI, to share why open collaboration is essential to shaping the next era of AI and how small, specialized models are driving a more accessible future.

Read more: How I use AI agents to automate my workflow and save hours

Sriram believes the future of AI depends on community intelligence as much as artificial intelligence. In his keynote, he emphasized that meaningful progress only happens when core technologies are developed in the open, shaped, and tested by a global community. That philosophy inspired IBM’s decision to contribute three major projects, Data Prep Kit, Docling, and BeeAI, to the Linux Foundation, ensuring developers everywhere can build, scale, and collaborate without barriers.

Each project plays a key role in the AI development pipeline. Data Prep Kit helps developers clean and prepare massive unstructured datasets for modeling, Docling converts complex documents like PDFs and PowerPoints into machine-readable formats, and BeeAI provides an open framework for building and orchestrating AI agents. Together, they form a foundation for building transparent, interoperable systems, and for inviting developers to experiment, extend, and innovate openly.

Sriram also shared why he believes “small is the new big” in AI. As models become more efficient, developers can achieve advanced results with smaller, fit-for-purpose LLMs that run on standard hardware. This shift, he said, makes AI more inclusive by reducing the need for massive infrastructure and promoting innovation at every level of expertise. With IBM’s Granite models released under the Apache 2.0 license, developers gain freedom to adapt and extend tools that are truly open.

Transparency is another cornerstone of IBM’s approach. By publishing detailed documentation about its training data, including a software bill of materials for AI, IBM is setting a higher standard for accountability in open source AI. Projects like Data Prep Kit allow developers to follow the same data-cleaning “recipes,” creating a shared foundation of trust and reproducibility across the community.

Key takeaways

• AI needs community: Open collaboration ensures that AI evolves responsibly and benefits everyone.
• Small is the new big: Fit-for-purpose models make AI development faster, cheaper, and more accessible.
• Transparency builds trust: Sharing datasets, documentation, and tools openly helps the entire ecosystem grow stronger.

Conclusion

Sriram reminds developers that open source isn’t just about code, it’s about community and connection. By contributing tools, models, and ideas in the open, we make AI more transparent, inclusive, and practical for everyone. The next wave of innovation, he says, will come from communities that build together.

More from We Love Open Source

• Want to get into AI? Start with this.
• Deep dive into the Model Context Protocol
• The secret skill every developer needs to succeed with AI today
• How I use AI agents to automate my workflow and save hours
• Skip the crowded job hunt: Find your tribe instead