LegalOn tunes Codex usage, estimates 65% cost cut
What happened
On October 8, OpenAI News reported that LegalOn cut its estimated daily cost by about 65% versus using GPT-5.5, while keeping development speed, by picking models per task and adjusting budgets. GPT-6 Luna handles well-specified code implementation, GPT-6.1 Sol handles routine design and analysis, and GPT-6 Astra handles complex design. The team also limits Fast mode by default and combines parallel task execution with tiered budget caps. The headline says cost was "halved"; the body gives the roughly 65% estimated daily reduction.
Written by AI from the coverage · updated 1 hour ago
Coverage
Follow the reports to see the story from different sides.
- OpenAI NewsLegalOn halves Codex costs while maintaining development speed
LegalOn 通过按任务选择模型和调整预算，在保持开发速度的同时，将预估每日成本较使用 GPT‑5.5 时降低约 65%。团队用 GPT‑6 Luna 处理明确需求的代码实现、GPT‑6.1 Sol 处理常规设计与分析、GPT‑6 Astra 处理复杂设计，并默认限制 Fast mode，结合并行执行任务和分层预算上限控制支出。
Heat over time
Not enough continuous observations to draw a trend yet.