SpaceX 财报数字挺好，股价怎么还跌了？ 最近开始买美股后，我第一次认真打开 SpaceX 的财报。营收涨了，利润也涨了，财报发布后股价却跌了。 单看每个数字我都认识，放到一起就不知道这份财报到底算好还是不好。股价跌，是公司真的出了问题，还是大家原本期待得太高，我也分不清。 我把 SpaceX最新一季的业绩公告、电话会文字稿和财报发布后的市场资料一起丢给 Ling-3.0-flash-Fin，让它帮我理一遍：SPCX 这季经营得怎么样，主要数字都在增长，股价为什么还会跌。公司自己公布的内容和市场对股价的解释，要分开写清楚。 买卖还是我自己决定。我需要 AI 做的，是把这些看不明白的材料读完，把关键数字、公司的解释和市场的反应整理到一起。哪个地方还想细看，我再顺着它给出的材料继续问。 Ling-3.0-flash-Fin 是蚂蚁百灵首个金融增强模型。它沿用 Ling-3.0-flash 的架构和长上下文能力，保持 124B 总参数、5.1B 激活参数，在此基础上继续做了金融语料预训练和领域后训练。 它主要处理年报、财务工作簿和多份研究材料。在投研场景里，可以用于信息检索、研究推理、估值建模和报告写作。 面向金融从业者与开发者，Ling-3.0-flash-Fin 将在 OpenRouter 开启为期一个月的限时免费 API 调用，方便用户在金融研究、数据分析与应用开发等场景中快速体验和验证模型能力。 OpenRouter 体验地址： 本文仅为模型体验记录，不构成投资建议。

High expectations vs. actual guidance causes 90% of post-earnings price drops. Using Ling-3.0-flash-Fin to quickly digest earnings calls and extract company explanations alongside market reactions is a massive workflow upgrade

Revenue up, profit up, stock down. That gap is exactly why a filing dump beats a vibes take.

That gap between good numbers and a falling stock is usually guidance or expectations, so having it separate company explanation from market reaction sounds way more useful than another summary that just says beat estimates.

Company explanation in one pile, market reaction in another. Do not mash them.

好数字不等于好股价。拆开公司解释和市场反应，比再看一份“超预期”摘要有用。

The project team is really generous, offering a free month with a time limit—this is a huge vote of confidence in their own product

对对对，你说的没错

First serious read of a print after starting US stocks is a very normal confusion.

I would use it the same way: read the pack, then I decide.

This is where AI can actually add value to investment research not by telling you whether to buy or sell, but by separating the company’s fundamentals from market expectations. A strong quarter and a falling stock price can both be true when expectations were even higher.

投资这家公司，我觉得最好的策略就是一部分作为底舱永远不动，长期持有，另外一部分资金用来滚动操作，尽量少做空。

This is exactly where AI can make financial research easier. The numbers tell you what happened, but AI helps connect the company’s results and market reaction in one place.

Beat versus miss the prior story is often the whole price move.

This is the part of investing that gets overlooked. A company can report strong numbers and still disappoint the market if those numbers fall short of what investors had already priced in. Separating business performance from market expectations is where the analysis gets interesting.

It would be so much more convenient if the cited materials could be marked out in sequence then when double-checking the numbers later, I wouldn't have to spend ages flipping through the documents again.

Indeed, even after seeing the numbers, I was still completely confused. This method of separating the company's official statements from market interpretations is particularly useful. I'll try it out for a free month first.

营收暴涨股价却跌，原来是高增长+高烧钱+稀释三重杀。Ling-3.0-flash-Fin把复杂材料拆得这么清楚，这个免费月必须冲一下。

冲一波！

Yeah, this is actually a good way to look at it. Sometimes a strong report still isn’t enough if the market was expecting even more.

Ling-3.0-Flash-Fin via WorkBuddy is worth playing, industry reports are easy now

我恨space X，把我套树上了，要是早点知道这个金融增强模型就好了

被套的可多了

这个大模型需要去思考并理解人类对于 SpaceX 这家企业所承载的梦想。目前这些公司还处于“试错率”的阶段，营收很难支撑起这个事实。 因此，模型要能够理解人类的情感与梦想，在估值框架里给出更多的想象空间。所以，把这些资讯提供给大模型，由它来帮我们推导出一个结果也颇为重要。

是的，它其实也有跟我讲过这样类似的结论

Ling-3.0-flash-Fin 沿用 124B/5.1B 激活、再加金融语料和后训练，听起来适合年报和工作簿，但 SpaceX 这类材料大量是电话会口径、合同披露和二手研究，检索和摘要能帮忙，估值建模还是要看它有没有把假设写明白。

I like the separation between what SpaceX actually reported and what the market thinks it means. That distinction gets lost so easily when a stock moves after earnings.

我恨space X，把我套树上了，要是早点知道这个金融增强模型就好了

Revenue surged but the stock price dropped anyway turns out it was a triple hit from high growth + heavy cash burn with dilution. Ling-3.0-flash-Fin broke down such complex material this clearly definitely gotta take advantage of this free month.

能把引用材料顺着标出来就太省心了，回头核对数字时不用再翻半天文件

有道理

标出来就舒服多了

revenue growth can look great while expectations quietly move the goalposts

非上市公司的财务材料本来就难核对，这类测试反而更看出模型对材料可靠度的处理。照这个思路我也想找个标的试试。

Good numbers ≠ up stock. Love this use case, let the AI separate the company's actual results from the market's reaction so you can figure out if it was fundamentals or just high expectations.

卧槽，懂了营收涨利润涨，股价反而跌，往往就是博主说的‘市场预期被拔得太高’。看来垂直模型确实更懂美股的套路。趁着OpenRouter上还有限时免费体验，高低得去把我持仓的几只科技股季度材料丢进去跑遍，感谢硬核评测分享，已收藏

市场预期被拔得太高 h哈哈

This is a useful way to look at a print where the numbers went up and the stock still dropped.

😯，金融增强模型！第一次听说，牛逼！

有想法了么？

This is where context matters more than headline numbers.

This is a really practical way to use a finance model, because rising revenue and a falling stock can both be true and still need different explanations.

A masterclass in cutting through market noise. Relying on specialized financial models to bridge the gap between complex earnings reports and actual market behavior is a game changer for retail investors.

这种场景我会更愿意交给Ling-3.0-flash-Fin 先把公司经营和市场预期拆开不然很容易看到营收利润都涨就直接判断成利好

投研场景里用于信息检索研究推理估值建模和报告写作，比单纯问买不买更有实际价值，因为它帮你整理材料而不是替你下结论。