On this page
We built leaderboards this week, and the first thing I learned is that ranking code is mostly an argument about who is allowed in the room.
If you rank every repo we've scored, the top is full of tiny projects with a perfect 10. 479 public repos score exactly 10.00, and 450 of them have fewer than 1,000 lines of code. That is technically correct and completely useless, a bit like ranking restaurants by how few dishes they have burnt.
So the leaderboards only rank repos with some weight to them: public, at least 500 GitHub stars, at least 100 files, not a fork, not archived, written mainly in one of the 13 languages we parse fully, and scored on their newest index. Today that's 147 repos, with a median score of 6.97. For anyone new to it, repowise's health score (1 to 10) rates how likely a file is to cause bugs and how hard it is to change, using static checks plus git history; the repo score is the average across files, weighted by lines of code. Every number below links to the repo's page, where you can see the files behind it.
One change since we first drafted this: the scores now count code files only. Our engine used to score JSON, YAML and Markdown files too, and they always got a perfect 10, which flattered any repo with a lot of translation files or docs. We re-scored everything on 6 October, and some well-known repos dropped by more than a point.
Each language at a glance
A language gets a "healthiest" only if it has at least 5 ranked repos.
| Language | Repos ranked | Median health | Healthiest |
|---|---|---|---|
| C# | 8 | 8.71 | MassTransit 9.3 |
| Java | 11 | 7.44 | low-level-design-course 8.4 (measured on 20 May) |
| TypeScript | 37 | 7.40 | ghostery/adblocker 8.9 |
| C++ | 5 | 7.34 | Deep Lake 8.0 |
| Python | 44 | 6.95 | Litestar 8.5 |
| Go | 19 | 6.80 | KEDA HTTP add-on 7.9 (measured on 3 June) |
| JavaScript | 12 | 6.53 | AlaSQL 9.1 (measured on 14 July) |
| Rust | 10 | 5.77 | Ratatui 7.8 |
Scroll the table sideways to see every column.
The language medians are the weakest numbers on this page. Five C++ repos and eight C# repos are tiny samples, so the medians say more about which repos people asked us to index than about the languages. The per-language lists below are more fun, and each one has its own leaderboard page.
The top of the whole list
The healthiest repo in the pool is MassTransit, a distributed messaging framework for .NET, at 9.3 across 5,533 files, which makes it #1 of 5,000+ indexed repos. For a framework whose whole job is to be complicated on your behalf, that is a little unreasonable. Archived repos are left out of the ranking, which is why a retired Microsoft sample app with a slightly higher score isn't on this list.
C# holds four of the top seven places. Second is Blazor.Diagrams at 9.3 across 295 files (#2 of 5,000+ indexed repos), and third is AlaSQL, the JavaScript leader, which comes up again below.
Python
Python has the most repos in the pool, 44, and a median of 6.95. Litestar leads at 8.5 across 1,167 files (#10 of 5,000+ indexed repos), and FastAPI sits right behind it at 8.4 across 2,927 files (#13 of 5,000+ indexed repos). Both were indexed this month. Third is camply, a tool that finds campsites at sold-out campgrounds, at 8.3 across 170 files (measured on 11 July). Someone clearly cares about that code, and probably about camping.
In the interest of full disclosure, our own repo is fifth on the Python list. We did not build the leaderboard for that, but I did check. The full list is on the Python leaderboard.
TypeScript
TypeScript has 37 repos and a median of 7.40. Ghostery's adblocker leads at 8.9 across 215 files, then flowsint at 8.5 (measured on 1 August) and shadcn/ui at 8.3 across 5,069 files (#14 of 5,000+ indexed repos).
The one I keep looking at is Grafana, eleventh on the TypeScript list at 7.9 with 18,146 files. All ten repos above it have fewer than 6,000 files, and Grafana is still in the top third of the language. More on the TypeScript leaderboard.
Go
Go has 19 repos and a median of 6.80. KEDA's HTTP add-on leads at 7.9 (measured on 3 June), then listmonk at 7.4 (measured on 30 June) and Gitea at 7.4 across 3,158 files (measured on 17 July). After the leader, the next four sit within about a tenth of a point of each other, and Go code tends to look like other Go code, which I assume is the point of Go. The rest are on the Go leaderboard.
JavaScript, Java, Rust, C# and C++
JavaScript has 12 repos. AlaSQL, a SQL database written in JavaScript, leads at 9.1 (measured on 14 July), then Ace, the code editor, at 8.4 across 1,245 files, and Fastify at 7.7. See the JavaScript leaderboard.
Java has 11 repos, and the top spot goes to low-level-design-course, the code that goes with a video course on low-level design, at 8.4 (measured on 20 May). Code written to teach good design scoring well is either very reassuring or very suspicious, and I'm choosing reassuring. Next come Java PathFinder at 7.7 (measured on 30 July) and Keycloak at 7.7 across 10,414 files, the biggest repo near the top of any list here. See the Java leaderboard.
Rust has 10 repos and the lowest median, 5.77. Ratatui leads at 7.8, then Vortex at 7.5 and Bevy at 6.7. See the Rust leaderboard.
C# has 8 repos and by far the highest median, 8.71: MassTransit, Blazor.Diagrams and supabase-csharp at 9.0 take the top three. See the C# leaderboard.
C++ reached five ranked repos in this refresh, just enough for a list. Deep Lake leads at 8.0 across 856 files, then Coinbase's cb-mpc at 7.8 (measured on 17 July). See the C++ leaderboard.
A language with one repo
Ruby has only 1 ranked repo, so it doesn't get a "healthiest" yet, but that one repo is worth a mention. Chatwoot, the open-source support desk, scores 8.9 across 7,989 files (#6 of 5,000+ indexed repos), and it is the biggest repo in the top ten. It used to score 9.6, back when its 640,000 lines of JSON translation files counted as perfect code, and it still sits near the top without them.
Limits of the score
- Size matters a lot. Across all scored public repos, median health falls from 9.91 under 1,000 lines of code to 6.79 between 100,000 and 999,999 lines, so a big repo at 7.9 is doing something a small repo at 8.9 doesn't have to.
- Each score comes from the repo's newest index, measured between May and October 2026. Where the index is more than 60 days old, I've written the date next to it, and some of these repos have moved since.
- The score reads structure and git history: complexity, function size, duplication, churn and similar markers. Whether the product is good, the tests are thorough or the maintainers are kind is outside what it measures, and those matter more.
- Primary language is GitHub's label, so a TypeScript repo with a Go backend is ranked as TypeScript.
If you maintain one of these and something looks wrong, tell us and we'll check it. If your repo isn't indexed yet, you can paste its GitHub URL into repowise.dev to see its score and where it would sit, and the same page lets you add a README badge if you like the number. If you don't like the number, the page also shows which files pulled it down, which is less fun but more useful.
The full lists, updated as we index more repos, are on the leaderboards. Nobody has burnt a dish there yet.
How we measured
- Who's ranked: public repos on repowise.dev with at least 500 GitHub stars and at least 100 files, not a fork, not excluded by hand, with a primary language among the 13 we parse fully (Python, TypeScript, JavaScript, Svelte, Vue, Java, Kotlin, Go, Rust, C++, C#, Scala and Ruby), and a health score on their newest ready index. Repos indexed under two owners count once. Archived repos are left out. That gives 147 ranked repos out of 5,163 repos with a ready index, which is where "5,000+" comes from.
- The score: as defined at the top, counting code files only since the 6 October re-scoring. Method: what is code health.
- Language: GitHub's primary language for the repo. A language needs at least 5 ranked repos to get a "healthiest".
- Rank: position in the 147-repo pool by score (ties broken by stars), written as "out of 5,000+ indexed repos" because that's the set every repo was compared against. Ranks are from the 7 October 2026 ranking refresh.