OpenRouter 发布自成立以来对 provider 路由器的最大改进，新算法不再主要按价格排序，而是按请求实际成本下的推理质量分配流量，让请求保持低价的同时更快更可靠。官方对比数据显示，新算法下 TTFT p50 从 1,007ms 降至 837ms、p95 从 4,601ms 降至 3,733ms，吞吐 p50 从 62.4 tok/s 升至 72.6 tok/s，整请求 p50 从 4.00s 降至 3.24s；代价是每千次请求成本从 $1.25 升至 $1.32、每百万 token 从 $0.301 升至 $0.324。失败重试比例从 10.4% 降至 6.9%，命中至少一次 429 的比例从 7.1% 降至 3.8%。
We just shipped our biggest improvement to our provider router since the founding of OpenRouter.
Before, we prioritized mostly based on price. The new router now rewards the best quality inference for the actual cost of a request, so requests stay cheap but are also fast and reliable.
Providers that deliver more value automatically win more traffic, so every provider has a reason to keep improving. And every time a provider gets better, so does your inference on OpenRouter.
来源：OpenRouter · x.com