Gloo has introduced Gloo Code, a new agentic development capability within Gloo AI Studio, as the company seeks to address one of the emerging constraints in AI-assisted software development: the cost and efficiency of model usage. Rather than relying on a single model for every coding task, the platform combines purpose-built agents with different leading models according to workload and complexity.

The launch comes as AI coding becomes increasingly mainstream among developers. Gloo cites research showing that 84% of developers are using or planning to use AI coding tools this year. Yet widespread adoption creates a management problem for organizations, where variable token consumption can make technology budgets difficult to forecast and explain.

Benchmark Results Highlight Cost-Efficiency Strategy

Gloo said preliminary testing on Terminal-Bench 2.1 produced approximately 70% performance while costing about half as much as comparable frontier models. Based on the company’s comparison with published results, Gloo Code completed the benchmark at 58% to 67% lower cost than several leading models while delivering competitive accuracy.

Those figures are potentially significant because the economics of AI development increasingly depend not only on model intelligence but also on how efficiently that intelligence is deployed. Gloo’s approach attempts to shift the optimization decision away from individual developers by determining which model and agent configuration is appropriate for each task.

However, the company explicitly characterizes its benchmark figures as preliminary internal results. Performance depends on testing conditions, workloads and model configurations, meaning the reported cost advantage should be viewed as an early indication rather than a universal measure of superiority.

Enterprise Predictability Becomes a Competitive Feature

Gloo is positioning predictable spending and privacy alongside technical performance. The company says Gloo Code offers flexible plans for individuals and teams, allowing engineering leaders to manage AI usage without choosing between enterprise privacy and more predictable pricing.

This positioning reflects a broader shift in enterprise AI purchasing. As organizations move from experimentation toward production deployment, the question is increasingly whether AI systems can deliver measurable productivity improvements without creating uncontrolled infrastructure expenses. Tools that automate model selection and reduce unnecessary token consumption could therefore become increasingly valuable as AI workloads scale.

Hackathon Provides an Immediate Testing Ground

The timing of the launch also connects Gloo Code directly with the Gloo AI Hackathon 2026, whose 30-day virtual build window has begun ahead of an in-person finale in Boulder, Colorado, scheduled for October 6–8. More than 500 builders are expected to use the platform during the virtual development period.

That environment gives Gloo an opportunity to observe how developers use its agentic coding system across practical projects rather than controlled benchmark conditions. The company says Gloo Code is built on the same infrastructure supporting its broader platform ecosystem, including Gloo 360, Barna, Servant and Midwestern.

The key question going forward will be whether Gloo can translate its early cost-efficiency claims into consistent productivity and financial benefits across real-world enterprise workloads. Adoption among developers, the reliability of model routing and the ability to maintain performance as projects become more complex will determine whether token optimization becomes a meaningful competitive advantage. For businesses expanding their AI development footprint, the combination of performance, predictable costs and privacy could become as important as the underlying models themselves.