The cheapest model just won the routing vote. langchain's open recipe for picking models per task…
the cheapest model just won the routing vote. langchain's open recipe for picking models per task cut openswe's median thread cost 64 percent without hurting merge rates, per october reporting.
the smart part of the agent is now the dispatch.
Context
Verified event date: LangChain published How to Build a Model Router in the Harness on Oct 1, 2026. It says that in its experiments, model routing cut median cost per coding task by 64 percent compared with the baseline, with no noticeable change in quality, using a router built into the harness of Open SWE, its open-source coding agent.
Falkster's summary says the router picks one of three model tiers on the first message, and that it was tested in an A/B on LangChain's own engineers.
The 64 percent median cost cut and Open SWE match LangChain's post. LangChain says 'no noticeable change in quality'; 'merge rates' as the quality measure is not seen in the excerpts read, so unsupported here, not refuted. The post says 'per october reporting', which fits Oct 1. 'The dispatch is the smart part' is the author's.
'the smart part of the agent is now the dispatch' is the author's opinion.
Related work
- LangChain: How to build a model router in the harness ↗Oct 1, 2026.
- dev48: How to build a model router in the harness ↗Mirror of the article.
- Falkster: Label before you route ↗Analysis of the router.
Watch next
- Read LangChain's post for how it measured quality.
Sources
- LangChain, Oct 1, 2026langchain.com
- dev48, Oct 1, 2026dev48.org
- Falkster, Label before you routefalkster.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 11 October 2026 at 06:16 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →