What We Measured After Moving 100B Tokens per Week to Open-Weight Models

ito.ai reports measurements made after moving roughly 100 billion tokens per week to open-weight models. The supplied metadata confirms the scale and the migration, but does not include model names, infrastructure details, cost changes, latency, quality metrics, or failure rates. Those details are essential for evaluating whether the results generalize beyond the authors’ deployment, so the article should be read as an operational case study until its full evidence is available.