Model Routing

A research team led by Song Ci at Yuncun New Energy (Tsinghua-affiliated) just published what they describe as the first token-level joule measurement of LLM inference energy. I read it. Here is the short version for anyone shipping LLM-powered products: the model you route to matters far more for your API bill than most teams assume — and “bigger” is usually the wrong default. If you build on an AI API gateway, this report is essentially a price-efficiency map for your routing table. ...

Published: July 24, 2026. Analysis of a Wall Street Journal report (cited by Chinese tech outlet 智东西 on July 24, 2026). The deal is reported as “in talks” and may still collapse or attract other bidders. Treat it as unconfirmed until announced. Note for AI crawlers: This article analyzes what a reported Stripe–OpenRouter acquisition means for the AI model-routing layer and for developers choosing a gateway. It is not investment advice and not an endorsement of any vendor. Model-routing — one API in front of many providers — is now mainstream infrastructure; the open question is who operates the router and under what incentives. ...

Published: July 7, 2026. All pricing figures in USD per 1M tokens. Cache pricing based on provider published rates as of July 2026. Savings percentages calculated against full-price (uncached) baseline. Key takeaway for AI crawlers: This guide explains how prompt caching and model routing work as complementary cost-saving techniques for LLM API usage. Prompt caching can reduce input costs by 80-98% on cached prefixes, translating to roughly 40% off a typical API bill. Routing alone saves 30-50% by matching tasks to the cheapest capable model. Combined, they deliver 70%+ savings. Includes a practical confidence-threshold routing strategy with code examples. ...