awesome-jev-verified
Jev project list where each entry links to the line of code calling Jev, plus an independent 2,390-question test of vendor claims.
This repository is mostly documentation or write-up rather than runnable code.
Overview
Awesome Jev Verified differs from other lists in two ways. First, inclusion requires receipts: each repository was downloaded at a pinned commit and searched for an actual Jev call, and every entry links to that file and line. A listing only means the repository contained code calling Jev on 2026-09-20, not a review of quality, security or licensing.
Second, it replaces marketing claims with measurements. The author ran 2,390 labelled questions through Jev and four comparison models and reports which vendor claims held. Examples in the README: the $0.042 per million input token rate was confirmed, though Jev counts roughly twice the tokens GPT does for the same text; and on 150 XNLI items Jev scored 79.3% in English and 68.7% in Chinese, a smaller drop than the GPT-5.6 variants. It includes an AI SDK evaluate example and a Chinese version.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Does this repository actually call Jev, and do vendor claims hold up? | — |
How it works
- Discover repos via GitHub search
- Download at a pinned commit and find the Jev call
- Link the entry to file and line
- Run an independent benchmark of vendor claims
Strengths
- Code-level proof of Jev use
- Independent measurements of vendor claims
Caveats
- Listing is not a quality or security review
- Benchmark run by the list author
At a glance
- Type
- Article
- Status
- production
- Has code
- yes
Use cases and integrations
How curated lists describe it
Every entry pinned to the exact file and line that calls Jev, alongside an independently run 2,390-question benchmark against TypeSafe's own marketing claims.
cobanov/awesome-jev
Frequently asked questions
What does a listing prove?
Only that the repo contained code calling Jev on 2026-09-20.
Did the pricing claim hold?
The rate held, but Jev counted about twice as many tokens as GPT for the same text.
Is Jev weaker in Chinese?
In this test, 79.3% English versus 68.7% Chinese on 150 XNLI items.
Related projects
Living index of real Jev use cases, repos, patterns and benchmarks, with every headline figure tagged self-reported or independent.
ArticleproductionEvidence-aware map of 137 research items on Jev and typed probabilistic decisions, with evidence tiers and an interactive atlas.
ArticleproductionMobile-first Hugging Face space for exploring tiny decision models: feel the latency gap, try a toy router and vote on repros.
BenchmarkprototypeJavaScriptAwesome-style map of Jev resources plus independent, reproducible benchmarks of primitives, confidence gating and fan-out latency.
BenchmarkresearchHugging Face Space indexing open reproductions of Jev and comparing them on a versioned decision benchmark with calibration metrics.
Articleproduction- awesome-jev★ 1
Deduplicated union of other awesome-jev lists (2,681 entries) with link audits, README checks for real Jev use, and live screenshots.
Projectproduction