According to Latent Space, competition among frontier model companies and the increasing capability of open-weight models such as Kimi K3 and Qwen3.8-Max are driving demand for 'model routing' — systems that decide which model to use for a given task. The publication cites Stripe's reported acquisition of OpenRouter for over $7B as one signal of the trend, and says it is equally active inside enterprises. Glean, co-founded and led by former Google Distinguished Engineer Arvind Jain, focuses on deploying AI in large organizations. Latent Space reports Glean was last valued at $7.2B following a $150M Series F round in June, and reached $300 million in annual recurring revenue this year, described as a three-fold increase over 15 months. Jain said part of Glean's approach is selecting which model to use per task, or determining whether a large language model is needed at all. "A big goal of Glean is to avoid using LLMs for tasks where we don't need them," Jain told Latent Space, noting that some queries involve simple arithmetic that a calculator could handle. Jain described Glean's aim as providing "one really powerful personal co-worker" to enterprise employees, acting as a 'meta-harness' across leading LLMs. Glean announced its third-generation Glean Assistant last September.
- Glean reportedly reached $300M ARR, a three-fold increase over 15 months, per Latent Space
- Glean was last valued at $7.2B after a $150M Series F in June
- Stripe reportedly acquired model-routing platform OpenRouter for over $7B
- Jain says a goal is to avoid using LLMs for tasks that don't need them, such as simple arithmetic
What it means for you
Model routing means automatically sending each task to the cheapest model that can do it well — or to no AI model at all when a plain calculation or lookup would do. The core idea here is simple and sound: don't pay for a top-tier model to add two numbers. For most small businesses this is a background trend among big vendors, not something you need to build yourself, but the principle is worth stealing.
Try this
Look at how you're using AI this week and ask, task by task, whether you actually need a large model for it — or whether a spreadsheet formula, a search, or a cheaper model would give the same result for less money.
Who should care
Businesses running enough AI queries that the monthly bill is noticeable, and technical teams building products on top of multiple LLM providers.
Skip this if
You use one or two AI tools casually and your costs are trivial — routing is an optimization problem you don't have yet.
Sources: Latent Space — read the original