The Best Model Routing is Task Specific最佳模型路由是任务导向的
Model routing is so hot right now. In the last few 1.5 months, OpenRouter shipped 模型路由现在非常火。在过去的一个半月里，OpenRouter 推出了, a compound model that fans your prompt out to a panel of frontier models and synthesizes one answer; Cognition shipped ，这是一个复合模型，它将你的提示词分发给一组前沿模型，并综合出一个答案；Cognition 推出了, a harness that runs a frontier model and a cheap "sidekick" in parallel and switches between them with classifiers; Factory shipped a ，这是一个并行运行前沿模型和一个廉价“辅助模型”，并通过分类器在两者之间切换的架构；Factory 推出了一个 that holds ~99% of Claude Opus 4.7's pass rate on Terminal-Bench 2 at 20% lower cost; even Vercel now ships ，它在 Terminal-Bench 2 上保持了 Claude Opus 4.7 约 99% 的通过率，同时成本降低了 20%；甚至 Vercel 现在也在其 AI 网关中推出了 in its AI gateway. Recently 。最近， published an article showing how they were able to tune the harness and model mixture to create a spreadsheet agent with much higher accuracy and lower tokens compared to Claude for Excel. 发表了一篇文章，展示了他们如何通过调整架构和模型组合，创建了一个比 Claude for Excel 准确率更高且 token 消耗更低的电子表格智能体。
The pitch converges on a common pattern: nobody wants to pay frontier prices for every token.这些方案都指向一个共同模式：没人愿意为每一个 token 都支付前沿模型的昂贵价格。
I think the trend is real. To add on to this, I do think that the best routing is deeply task-specific, and the more narrowly you focus on a single workflow, the more alpha there is to exploit in terms of both accuracy and cost. You don't need frontier intelligence for every task — but figuring out 我认为这个趋势是真实的。此外，我深信最佳的路由方式是高度任务导向的。你对单一工作流的聚焦越窄，在准确性和成本方面可挖掘的 Alpha（超额收益）就越多。你并不需要为每个任务都使用前沿智能——但弄清楚which哪个 model clears the bar for 模型能够满足which task, subject to your cost/latency constraints, is a problem only someone obsessed with that task can solve.该任务的需求（在你的成本/延迟约束下），这是一个只有深耕该任务的人才能解决的问题。
Model intelligence is jagged, and what counts as "good enough" (and what you'd pay for the last few points of accuracy) depends entirely on the task. A router that doesn't know your task is guessing at both.模型智能是不均衡的，什么才算“足够好”（以及你愿意为最后那一点准确率支付多少代价）完全取决于任务本身。一个不了解你任务的路由器，在这两方面都只能靠猜。
The cost story, and why a generic router falls short成本逻辑，以及为什么通用路由器会力不从心
Frontier models are too expensive to run on every token at scale — a model like Fable can burn $600 an hour of inference at volume, and most work doesn't need that horsepower. The curve helps: near the top, cost drops sharply while quality barely moves, because the first work to leave the frontier model is the work a cheaper one handles just as well. Factory runs on that flat stretch and pulls out 20-25% of the cost while pass rate holds; Cognition's Devin Fusion, which declares "the age of using one model for all of your work is coming to an end," holds Fable 5-level performance at 35% lower cost, with 88% of its users' merged PRs now coming straight from the router.前沿模型在大规模处理每个 token 时过于昂贵——像 Fable 这样的模型在大规模使用时每小时推理成本可达 600 美元，而大多数工作并不需要这种算力。曲线说明了一切：在接近顶端时，成本急剧下降，而质量几乎没有变化，因为首先被从前沿模型中剥离出来的任务，往往是较廉价模型也能同样处理好的工作。Factory 正是运行在这一平缓地带，在保持通过率的同时降低了 20-25% 的成本；Cognition 的 Devin Fusion 宣称“使用单一模型处理所有工作的时代即将结束”，它在保持 Fable 5 级别性能的同时降低了 35% 的成本，且其用户 88% 的合并 PR 现在都直接来自该路由器。
OpenRouter is genuinely great at this, and I said as much when Fusion launched. But Fusion is an ensemble built to maximize quality on hard, open-ended questions where you'll pay for several completions because being wrong is expensive. Task-specific routing answers a different question: given OpenRouter 在这方面确实做得很好，Fusion 发布时我也这么说过。但 Fusion 是一个为了在困难、开放式问题上最大化质量而构建的集成系统，你愿意为此支付多次补全的费用，因为出错的代价很高。任务导向的路由解决的是另一个问题：针对this specific input这个特定的输入, what's the cheapest path that maximizes ，什么是最廉价且能最大化this task's该任务 quality bar? A generic gateway can route across providers and fall back when an endpoint degrades, but it can't know that page 3 of your scanned loan file is a dense table that needs a specialized vision model while pages 1-2 are plain text you should pull with cheap direct extraction. It has no model of your inputs, because it isn't in the business of any one task.质量标准的路径？通用网关可以在不同提供商之间路由，并在端点降级时进行切换，但它无法知道你扫描的贷款文件第 3 页是一个需要专门视觉模型处理的密集表格，而第 1-2 页是完全可以通过廉价直接提取处理的纯文本。它没有你输入内容的模型，因为它并不专注于任何特定任务。
The teams routing well go one workflow deep路由做得好的团队，都深耕于单一工作流
The coding agents are just the loudest example. Look at the teams shipping serious vertical AI products and you see the same pattern: each routes inside a single workflow, and each has gone deep enough on its task to know which model clears the bar.编码智能体只是最明显的例子。看看那些发布了严肃垂直 AI 产品的团队，你会发现同样的模式：每一个都在单一工作流内部进行路由，并且每一个都对自己的任务钻研得足够深，从而清楚哪个模型能达标。
Harvey, in legal, is the cleanest case. Their 法律领域的 Harvey 是最典型的案例。他们的 breaks a request into sub-tasks, picks a model for each, and synthesizes — "no single model is the best at everything," so they route high-volume Vault work to faster models like Sonnet 4.6 and Gemini 3 Flash where latency matters and quality is good enough, and they're 将请求拆分为子任务，为每个任务选择模型并进行综合——“没有哪个模型在所有方面都是最好的”，因此他们将大批量的 Vault 工作路由到 Sonnet 4.6 和 Gemini 3 Flash 等对延迟敏感且质量足够好的快速模型上，并且他们正在 to hit near-frontier legal performance for less.，以更低的成本达到接近前沿水平的法律处理能力。
Decagon does the same in support, and Jesse's framing is the sharpest version of the argument I've read: when a use case is new you want the smartest general model, but "once the use case is fully built out... general intelligence is overhead," and you want "the smallest, fastest model fine-tuned to do your specific thing extremely well." Decagon now runs Decagon 在客服领域也做了同样的事情，Jesse 的论点是我读过最犀利的：当用例是全新的，你需要最聪明的大模型，但“一旦用例完全成熟……通用智能就是一种冗余”，这时你需要的是“针对你的特定任务进行微调的、最小最快的模型”。Decagon 现在将其 on fine-tuned open models.运行在微调后的开源模型上。
Neither is reaching for a single model and hoping. Each is hillclimbing one task with a mix of models it understands deeply, and that understanding is the thing a generic router can't buy. It's the most concrete version of Sarah Guo's 他们都没有指望单一模型能解决所有问题。每一家都在通过深入理解的一系列模型来优化单一任务，而这种理解是通用路由器无法买到的。这是 Sarah Guo 提出的: the difficulty model and the eval that grades it are private ground truth, earned from real volume, that a smarter frontier model next quarter doesn't hand you. That's where the alpha is, and it compounds.概念的最具体体现：难度模型和评估它的标准是私有的事实真相，是从真实业务量中积累出来的，下个季度出现的更强前沿模型并不会直接带给你这些。这就是 Alpha 所在，而且它会产生复利效应。
Document AI: the gap is big, and it stays big文档 AI：差距巨大，且将持续存在
A lot of people assumed that frontier models would eat document parsing entirely. It didn't. On 很多人曾假设前沿模型会完全吞噬文档解析领域。但事实并非如此。在, the 上，即我们, frontier VLMs are great at visual understanding and terrible at layout — GPT-5-mini and Haiku score below 10% on visual grounding while specialized parsers land at 55-80%, and no single method tops all five dimensions. Gemini gains ~5 points going from minimal to high thinking at 4x the cost, and you burn vision tokens on text-heavy pages that never needed a VLM.中，前沿视觉语言模型（VLM）在视觉理解方面表现出色，但在布局理解上却很糟糕——GPT-5-mini 和 Haiku 在视觉定位上的得分低于 10%，而专门的解析器则能达到 55-80%，且没有任何一种单一方法在所有五个维度上都名列前茅。Gemini 从极简思维模式切换到深度思维模式，成本增加了 4 倍，得分仅提升约 5 分，而且你会因为在纯文本页面上消耗视觉 token 而浪费资源，这些页面根本不需要 VLM。
The right architecture for document OCR involves a complex "router" (or more accurately, a complex engine) with many pieces operating in conjunction that are carefully tuned for the specific data domain and task distribution. 文档 OCR 的正确架构涉及一个复杂的“路由器”（更准确地说是复杂的引擎），它由许多协同工作的组件组成，并针对特定的数据领域和任务分布进行了精心调整。 contains this under the hood: an agentic harness that auto-routes each page between frontier and specialized models with a self-improving document-complexity model, custom document engines that cut vision tokens 50-90% where text is all there is, specialized VLMs post-trained for tables and charts, and an agentic judge that validates the output. On the cost-accuracy plane it sits on the frontier — our agentic mode leads at 84.9% and wins 4 of 5 dimensions, our cost-effective mode lands near the top at roughly a third of a cent per page. 在底层包含了这一逻辑：一个智能体架构，通过自我改进的文档复杂性模型，自动在通用模型和专用模型之间路由每一页；定制的文档引擎，在纯文本页面上减少了 50-90% 的视觉 token 消耗；针对表格和图表进行后训练的专用 VLM；以及一个验证输出的智能体裁判。在成本-准确率平面上，它处于前沿水平——我们的智能体模式以 84.9% 的准确率领先，并在 5 个维度中赢得了 4 个，我们的高性价比模式以每页约三分之一美分的成本处于领先梯队。
There will always be a massive gap on the pareto curve for document understanding relative to the visual capabilities of frontier models. We have designed LlamaParse to always be the best at exploiting that gap. 在文档理解领域，相对于前沿模型的视觉能力，帕累托曲线上将始终存在巨大的差距。我们设计 LlamaParse 的初衷，就是始终成为利用这一差距的最佳方案。
Two layers, different jobs两个层级，不同职责
The two layers settle into different roles. Generic gateways like OpenRouter own broad, provider-level routing and ensembling (which models are up, which are cheapest this minute, when to convene a panel for a hard general question). The task-specific layer is where the excess returns live, and they accrue to whoever goes deepest on a workflow: software engineering to Factory and Cognition, legal to Harvey, support to Decagon. For documents, we intend it to be us. If you're routing frontier and open-source VLMs to squeeze every point of document accuracy out of every cent, 这两个层级各司其职。像 OpenRouter 这样的通用网关负责广泛的、提供商层面的路由和集成（哪些模型可用、此刻哪个最便宜、何时针对困难的通用问题召集专家组）。任务导向层才是超额收益所在，这些收益归属于那些在工作流中钻研最深的人：软件工程归 Factory 和 Cognition，法律归 Harvey，客服归 Decagon。对于文档领域，我们致力于成为那个深耕者。如果你正在路由前沿模型和开源 VLM，以求从每一分钱中榨取文档准确率的每一分提升，.
想发布自己的文章？想发布自己的文章？
升级为 Premium升级为 Premium
