NaceAI 发布 Drex Router:57-87 毫秒完成 LLM 路由决策
NaceAI 刚上线的 Drex Router 用 12B 的 Gemma 模型在 90 毫秒内帮你挑 LLM,价格只有 Jev Router 的十分之一。做多模型调用、想省 token 钱的开发者可以看看,它还带概率和理由输出,方便做降级兜底。
NaceAI 推出 Drex Router,一个放在 LLM 调用前的路由器。它基于 Gemma 构建的 12B 模型读取提示词和候选模型列表,约 57-87 毫秒返回选择结果,单次决策延迟低于 90 毫秒,几乎可插入任何 LLM 调用链路。
在 Nace 的 1,400 道常识考试题测试中,Drex Router 达到 87.6% 的路由准确率,成本为每 1,000 次查询 0.10 美元;对比的 Jev Router 准确率 86.4%,成本 1.09 美元,价格相差约 10 倍。
区别于只返回模型名的路由器,Drex Router 对每个候选模型输出一个概率值和一段简短理由,使用者能区分高置信选择和近乎随机的猜测,便于在路由不可靠时设置降级或人工兜底策略。
A model decision in under 90 ms is fast enough to put a router in front of almost any LLM call.
@NaceAI just launched Drex Router, and it reads the prompt and your list of models, then returns its pick in about 57 to 87 ms, using a 12B model built on Gemma.
In Nace’s tests on 1,400 general-knowledge exam questions, it scored 87.6% for $0.10 per 1,000 questions, while rival Jev Router scored 86.4% for $1.09.
A router that only returns a model name can’t tell you a confident pick from a coin flip. Drex Router returns a probability for every candidate plus a short reason,
This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Drex Router, NaceAI, Gemma, LLM Router.
Source attribution: curated from rohanpaul_ai (@rohanpaul_ai).