Doesn't really make sense as there are already open weight models that are close to frontier models, and the compute to run them isn't outrageously expensive, and every indication that it will only get cheaper over time as has been the strong trend in terms of $/outcome.