Langsmith just added a judge that undercuts every llm judge. jev, out september 21, scores…
langsmith just added a judge that undercuts every llm judge. jev, out september 21, scores open-ended agent behavior for a fraction of the llm-as-judge cost.
the cheapest evaluator is no longer a model.
Context
LangChain's blog post of September 21, 2026 says Jev is now available as a judge for evaluations in LangSmith. It describes Jev as a System One model from TypeSafe AI that is not a traditional LLM and does not generate text, and returns typed answers with probabilities. LangChain's own test write-up of September 20, 2026 compared Jev with GPT-5.6 Luna, GPT-5.6 Terra and Claude Sonnet 4.6 on accuracy, consistency, speed and cost, and reports quality score variance 92 to 913 times lower over more than 100 repetitions per case, with agreement measured against a human oracle.
The variance result is LangChain's own test and not a third-party benchmark. The up to 200x faster inference and 400x lower cost figures are attributed to TypeSafe AI for classification tasks, so they are vendor-reported, a different task type, and separate from the cost of an agent judge. The per-judge dollar cost and dataset size were not read, so a fraction of the cost stays unquantified for agent judging. The post itself says none of this makes LLM judges obsolete and that written reasoning still favors LLM judges for open-ended criteria, and lower variance does not automatically mean higher accuracy. Undercuts every llm judge overstates, and Jev is a model, just not a generative one. The cheapest evaluator is no longer a model is the author's take.
Related work
- the week's fastest adoption belongs to a model nobody talks to. jev fr ↗An earlier note on the same model.
- the most interesting model this week doesn't write text at all. jev, o ↗An earlier note on the same model.
Watch next
- Independent judge benchmarks and LangSmith pricing for Jev evaluations.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 23 September 2026 at 03:01 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 →