OpenAI自研Jalapeño芯片2026年底部署
Semianalysis on OpenAI's new Jalapeño chips that will be deployed inside its own…
Semianalysis on OpenAI's new Jalapeño chips that will be deployed inside its own compute by the end of 2026.
Semianalysis 对 OpenAI 新型 Jalapeño 芯片的分析,该芯片将于 2026 年底部署在其自有计算设施中。
"Jalapeño smokes every other chip. All this is done without Multi Token Prediction (MTP), while the other chips on the chart are the best performing configs of each respective SKU, all with MTP"
“Jalapeño 在性能上远超其他所有芯片。这一切都是在没有多令牌预测(MTP)的情况下实现的,而图表中的其他芯片则是各自 SKU 的最佳配置,且均采用了 MTP。”
Jalapeño shifts the entire latency–efficiency Pareto frontier upward: at ~100 tok/s/user it delivers ~11M tok/s/MW, roughly 2× the best Blackwell-class configs at comparable interactivity.
Jalapeño 将整个延迟-效率帕累托前沿向上推移:在约 100 令牌/秒/用户的情况下,它提供约 1100 万令牌/秒/兆瓦,大约是同等交互性下最佳 Blackwell 级配置的 2 倍。
The kicker: that’s STP (Single-Token Prediction) with no speculative decoding, while the competing curves are already using MTP.
关键在于:这是采用单令牌预测(STP)且无推测解码的结果,而竞争曲线已经使用了 MTP。
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力