Silvera team roll out model router, 29% lower calculation cost

The Silvia team announced the launch of the latest model router, which would reduce costing by as much as 29 per cent, while improving user experience. By analysing the top 500 characters of the query, the model router predicts the level of effort required to calculate, thus moving the high-demand query path to the most powerful model, while the low-demand query route to the more appropriate model. Silvia also announced that its AI tax products had become the most accurate products on the market, going beyond OpenAI, Anthropic, Google and XAI。