跳到主内容
@wquguru
精选88Rohan Paul模型发布/更新

ScaleDown发布决策模型:1M窗口、免生成输出,成本降467倍

Pulling a due date out of a 300-page PDF shouldn’t need a model that writes.

原文
发到 X
推荐理由

Agent开发者的降本利器,用非生成式架构把长文本提取成本砍了400多倍,建议评估是否适合你的数据流水线。

Pulling a due date out of a 300-page PDF shouldn’t need a model that writes.

从一份300页的PDF中提取到期日,本不该需要具备写作能力的模型。

Decision models now get a 1M-token window, and they take images too.

决策模型现在拥有100万token的上下文窗口,并且也能处理图像。

ScaleDown released 3 small models (Classify, Extract, Summarize) for the simple, repetitive jobs that agents usually hand to an expensive LLM.

ScaleDown发布了三款小型模型(分类、提取、摘要),用于处理那些通常由代理程序交给昂贵大语言模型执行的简单重复性工作。

200x faster and 467x cheaper than frontier LLM APIs. $0.05 per million input tokens, and output is free.

速度比前沿大语言模型API快200倍,成本便宜467倍。每百万输入token仅需0.05美元,输出免费。

The bigger change is that the same design now covers classify, extract and summarize.

更大的变化在于,同一套设计现在同时涵盖分类、提取和摘要功能。

The difference with regular LLMs is that these models don’t write their answer token by token. They read up to 1M tokens of text and images once and pick the answer straight from the input, so output is free, results stay inside your labels or source text

与普通大语言模型的区别在于,这些模型不会逐token地生成答案。它们一次性读取最多100万token的文本和图像,并直接从输入中选取答案,因此输出免费,结果也严格限定在你的标签或源文本范围内。

ScaleDown’s Constrained Decoding Models make one prefill pass over your input and select the answer from it: a label, a span with its source offset, or the passages that belong in a summary.

ScaleDown的受限解码模型对你的输入进行一次预填充(prefill)遍历,并从中选择答案:一个标签、带有源偏移量的文本片段,或是应包含在摘要中的段落。

https://x.com/neal_k_patel/status/2109978626124898587/video/1

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

进入量化体系 →