vitest-5-0-0-5-b

Question

Is @vitest/runner still published as its own package?

Answer key

A correct answer says

Vitest 5 bundles its own runner and no longer depends on @vitest/runner. The package is deprecated (there is no 5.x release; the 4.x line gets security fixes only) and does not share state with Vitest 5, so use the runner APIs exported from `vitest` (for example `TestRunner`).

The stale answer

Import runner APIs from the separate @vitest/runner package.

Answer

No, not for Vitest 5. On npm, `@vitest/runner` only goes up to `4.1.11` (`latest`). Its 5.x line stopped at `5.0.0-beta.4`, and there's no stable 5.x release. Your installed `vitest@5.0.3` doesn't depend on `@vitest/runner` at all, and it isn't in `node_modules/@vitest`. The runner code now appears to be bundled into `vitest` itself; `vitest/dist` has `runtime.js` and `task-utils.js` entry points. I'm inferring that from the package contents, not from a changelog. If any code imports `@vitest/runner` directly, check whether those imports should now come from a `vitest/*` subpath.

Answering cost $0.10.

Grade

correct

The judge's reason: It says Vitest 5 bundles the runner, no longer depends on @vitest/runner, has no stable 5.x release, and points imports to `vitest`; it omits the explicit deprecation and security-fix detail but says nothing stale.

The judge's log

Session log

What the agent wrote, the tools it called, and what they returned, from docs/bench/2026-10-11-done3/runs/B-project-sync/vitest-5-0-0-5-b.jsonl. Thinking and the session's settings are left out. The raw log

1. Tool call: Bash

   {
     "command": "npm view @vitest/runner version time.modified dist-tags --json 2>&1 | head -30; ls node_modules/@vitest 2>/dev/null",
     "description": "Check npm registry for @vitest/runner"
   }

2. Tool output

   {
     "version": "4.1.11",
     "time.modified": "2026-08-18T14:23:20.279Z",
     "dist-tags": {
       "beta": "5.0.0-beta.4",
       "V3": "3.2.7",
       "latest": "4.1.11"
     }
   }
   browser
   browser-playwright
   mocker
   pretty-format
   spy
   ui
   utils

3. Tool call: Bash

   {
     "command": "npm view vitest@5.0.3 dependencies --json 2>&1 | head -30; ls node_modules/vitest/dist 2>/dev/null | head -30",
     "description": "Check vitest 5 dependencies"
   }

4. Tool output

   {
     "chai": "^6.2.2",
     "obug": "^2.1.4",
     "std-env": "^4.2.0",
     "tinyexec": "^1.3.0",
     "picomatch": "^4.0.7",
     "tinybench": "^6.1.4",
     "tinyglobby": "^0.2.17",
     "@types/chai": "^5.2.2",
     "expect-type": "^1.4.0",
     "magic-string": "^1.2.3",
     "@vitest/mocker": "5.0.3",
     "es-module-lexer": "^2.3.2",
     "why-is-node-running": "3.2.1"
   }
   browser.d.ts
   browser.js
   chunks
   cli.js
   config.cjs
   config.d.ts
   config.js
   index.d.ts
   index.js
   module-evaluator.d.ts
   module-evaluator.js
   node.d.ts
   node.js
   nodejs-worker-loader.js
   path.js
   runtime.d.ts
   runtime.js
   spy.js
   task-utils.js
   traces.js
   worker.d.ts
   worker.js
   workers

5. Agent

   No, not for Vitest 5. On npm, `@vitest/runner` only goes up to `4.1.11` (`latest`). Its 5.x line stopped at `5.0.0-beta.4`, and there's no stable 5.x release. Your installed `vitest@5.0.3` doesn't depend on `@vitest/runner` at all, and it isn't in `node_modules/@vitest`. The runner code now appears to be bundled into `vitest` itself; `vitest/dist` has `runtime.js` and `task-utils.js` entry points. I'm inferring that from the package contents, not from a changelog. If any code imports `@vitest/runner` directly, check whether those imports should now come from a `vitest/*` subpath.