# LegalOn Achieves 65% Cost Reduction with Tiered Codex Model > By implementing a strategic three-pronged approach to model utilization, LegalOn Technologies drastically reduced its estimated Codex costs, redefining budget management in AI operations. **Source**: fourweekmba.com | **Published**: 2026-10-10 | **Type**: case_study ## Key Facts - LegalOn cut estimated daily Codex costs by 65%, showcasing effective cost management strategies. - Transitioning to a tiered model approach reveals vulnerabilities in previous unlimited access policies. - Budget caps encourage efficiency, indicating a strategic shift towards sustainable AI usage and growth. - The focus on cost per feature release highlights the need for clearer financial metrics in AI projects. - LegalOn's model routing strategy provides a competitive advantage in optimizing resource allocation. ## Summary \## Summary LegalOn Technologies, a Tokyo-based legal-AI company, faced the challenge of managing escalating costs associated with AI model usage. By implementing a structured approach to model selection, feature restrictions, and budget management, the company achieved a reduction of approximately 65% in estimated daily Codex costs compared to GPT-5.5 while maintaining development speed. \## Background LegalOn Technologies operates in the legal technology industry and is known for its AI-driven solutions. Prior to the deployment of the new model management strategy, the company provided its developers with unlimited access to the GPT-5.5 model in Fast mode, which led to concerns about exceeding their annual budget due to high operational costs. \## Challenge The primary issue was the unsustainable cost of using high-performance AI models without restrictions. LegalOn recognized that continuing this approach would inevitably exceed their budget, prompting the need for a more efficient model usage strategy. \## Solution LegalOn's AI-powered Development Center of Excellence (AID CoE) developed guidelines for model selection, allowing teams to choose the appropriate model based on task complexity. They implemented a tiered model system with three different models: GPT-6 Luna for straightforward tasks, GPT-6.1 Sol for standard design and analysis, and GPT-6 Astra for complex tasks. Additionally, they turned off Fast mode by default and set monthly usage limits for departments and individuals, which were monitored and adjusted as needed. \## Results The implementation of the new model management strategy resulted in a 65% reduction in estimated daily costs compared to the previous use of GPT-5.5. This significant cost saving was achieved while maintaining the speed of development, although specific metrics on development speed were not provided. \## Key Insights 1. Implementing a structured model selection process can lead to substantial cost savings in AI operations. 2. Setting usage limits and monitoring model performance can help manage budgets effectively without sacrificing development speed. 3. Balancing cost efficiency with the need for rapid development is crucial, especially for new business initiatives. \## Customer Testimonial “Excessive restrictions through rules and budgets can undermine an organization’s momentum. What we need is a flexible operating model that effectively balances risk control with the speed teams need.” — Yuta Tokitake, Senior Engineering Manager at LegalOn Technologies. ## Entities - **Companies**: OpenAI, LegalOn Technologies - **Products**: Codex, GPT-5.5, GPT-6 Luna, GPT-6.1 Sol, GPT-6 Astra - **Technologies**: AI, model routing - **People**: Yuta Tokitake ## Key Concepts cost reduction, model selection, feature restrictions, budget management, development speed, AI-powered Development CoE, task complexity, flexible operating model ## Definitions - **model routing**: A process of directing requests to the most cost-effective model capable of handling them. - **AID CoE**: AI-powered Development Center of Excellence, which provides guidelines for model selection and monitors usage. - **Fast mode**: A high-performance operational mode for AI models that can be restricted to manage costs. - **budget caps**: Limits set on spending for departments or individuals to control costs associated with AI usage. - **feature release**: The deployment of a new feature in a software product, which can be tracked for associated AI costs. ## Use Cases - cost efficiency improvements in AI development - model selection for task-specific applications - budget management for AI resources - monitoring AI usage across departments - tracking AI costs per feature release - enhancing development speed while managing costs ## Frequently Asked Questions **How did LegalOn reduce its Codex costs?** LegalOn reduced its estimated daily Codex costs by approximately 65% by implementing model selection, feature restrictions, and budget management tailored to business stages. **What is model routing?** Model routing is the practice of directing requests to the most cost-effective AI model that can handle the task, ensuring efficient resource use. **What role does AID CoE play in LegalOn's strategy?** The AID CoE at LegalOn develops guidelines for model selection, tests and monitors model performance, and helps teams choose the appropriate model for their tasks. **What are the implications of turning off Fast mode by default?** Turning off Fast mode by default helps manage costs but raised concerns about development speed, prompting teams to run tasks in parallel to maintain performance. **What future metrics is LegalOn planning to implement?** LegalOn is working on a metric that links the customer value of each feature release to the AI costs incurred, aiming to make the total cost of a release more transparent. ## Links - [Read on Welcome.AI](https://welcome.ai/content/legalon-achieves-65-cost-reduction-with-tiered-codex-model) - [Original source](https://fourweekmba.com/ai-legalon-cut-estimated-codex-costs-65-by-matching-models/) - [LegalOn Technologies](https://welcome.ai/company/legalon-technologies): Featured company --- Source: Welcome.AI | https://welcome.ai/content/legalon-achieves-65-cost-reduction-with-tiered-codex-model