OpenAI宣布向所有客户广泛提供其决策API模型，正式进入快速、低成本决策模型市场的竞争 1。该模型通过优化推理性能，使客户能够用决策模型替代大语言模型以实现更高吞吐量 1。
OpenAI的决策API将推理延迟从1-3秒大幅降低至100-180毫秒 1。OpenAI API产品主管Nikunj Handa表示："决策API的重大突破在于速度，将响应时间从一到三秒缩短到100、120、180毫秒" 1。该模型基于ChatGPT Luna改进而成，在Box、Datadog、Ramp等客户初期试验中已取得成效 1。此外，决策API新增语音和视觉模式，并通过地理位置优化推理资源进一步降低延迟 1。
在性能竞争中，OpenAI的决策模型在AIM System One Index基准测试中排名第八，在Artificial Analysis Decision Index中排名第十四，而其主要竞争对手TypeSafe的Jev模型则分别排名第七和第三 1。TypeSafe的Jev模型已为该公司带来76亿美元的估值 1。
OpenAI has made its Decision API model widely available to all customers, marking an entry into the rapidly growing market for fast, cost-efficient decision models 1. The model competes directly with TypeSafe's Jev, which has driven the startup's valuation to $7.6 billion 1.
The Decision API achieves a significant performance improvement by reducing inference latency from 1-3 seconds to 100-180 milliseconds 1. According to Nikunj Handa, OpenAI's API product lead, "The big unlock with Decisions API has been the speed, shrinking things from one to three seconds to 100, 120, 180 milliseconds" 1. This acceleration enables customers to replace large language models with decision models to achieve higher throughput. Early customer pilots, including Box, Datadog, and Ramp, have demonstrated tangible benefits, with one case showing response time reduction from 1.4 seconds to 0.13 seconds 1.
In benchmark comparisons, Jev ranks seventh on the AIM System One Index and third on the Artificial Analysis Decision Index, while OpenAI's decision model ranks eighth and fourteenth respectively 1. The Decision API also introduces new speech and vision modalities, and optimizes inference resources through geographic distribution to further reduce latency 1.
评论
还没有评论，欢迎留下第一条。