A new AI model just proved that you don't need Google's computational firepower to give AI tools the ability to use other software. The breakthrough could reshape who gets access to advanced AI capabilities.
Cactus Compute released Needle, a 26-million parameter language model that can call external tools and APIs. The company created it by distilling Google's Gemini model — essentially teaching a much smaller AI to mimic the tool-calling behavior of its massive counterpart.
Tool calling lets AI models interact with other software, databases, and services. Instead of just generating text, these models can book appointments, query spreadsheets, or pull real-time data. Until now, this capability required expensive, large-scale models that only big tech companies could afford to run.
The technical achievement matters because of the dramatic size difference. Google's Gemini models contain hundreds of billions of parameters and require significant computing resources. Needle achieves similar tool-calling performance with 99% fewer parameters, making it cheap enough to run on standard business hardware.
This represents a broader shift in AI development. Large companies have dominated advanced AI capabilities by building ever-bigger models. But distillation techniques let smaller teams extract specific skills from these giants and package them into efficient, specialized tools.
Why Small Businesses Should Care
Tool-calling AI could automate workflows that currently require expensive custom software or manual processes. A small accounting firm could use it to automatically pull client data from multiple systems. A local retailer might connect it to inventory management and customer service platforms.
The cost difference is substantial. Running large models like Gemini through APIs can cost hundreds of dollars monthly for active businesses. Needle-sized models might run for under $50 per month on cloud services, or even locally on business computers.
This opens AI automation to companies that couldn't justify the expense of enterprise-grade solutions. Small businesses often need AI that can work with their existing tools — customer relationship management systems, accounting software, scheduling platforms. Needle-class models make that integration financially viable.
The Competitive Landscape Shifts
Other companies will likely follow this distillation approach. OpenAI, Anthropic, and Google have invested billions in large models. But if startups can extract their key capabilities into smaller, cheaper versions, the competitive advantage shifts toward execution and integration rather than raw computational power.
This could accelerate AI adoption across small and medium businesses. When advanced capabilities become affordable, more companies experiment with automation. That creates demand for AI integration services, custom solutions, and industry-specific tools.
What to Watch
The real test comes with real-world performance. Lab benchmarks don't always translate to reliable business use. Tool-calling AI needs to handle edge cases, error conditions, and integration challenges that crop up in actual workflows.
Companies should also watch how the major AI providers respond. They might accelerate their own small model development or adjust pricing to maintain market position.
The Bottom Line
Needle proves that advanced AI capabilities don't require massive budgets or enterprise contracts. Small businesses should start identifying workflow automation opportunities that seemed too expensive six months ago. The tools to make them affordable are arriving faster than expected.