Benchmark: search doctors on jameda with an AI agent
On jameda, Wield's recipe passed 9 of 9 doctor searches, with a median of 0.4 seconds and one request per run. A browser agent (Browser Use driving Claude Sonnet 5.5) also passed 9 of 9, with a median of 51 seconds and $0.15 per run. Both sides ran the same three specialty and city queries on 7 November 2026.
jameda Search doctors by specialty and city
Read-only, no loginPasses the daily check
Parameters (5)
| Metric | Wield recipe | Browser Use + Claude Sonnet 5.5 |
|---|---|---|
| Success | 100% | 100% |
| Median time | 447 ms | 51 s |
| Cost per run | $0 | $0.148 |
9 runs each, 7 November 2026. How we measure
The task
Each side ran the same three searches. Each search names a specialty and a city.
| Query | Parameters |
|---|---|
| q1 | specialty = Allgemeinmediziner, city = Hamburg |
| q2 | specialty = Zahnarzt, city = München |
| q3 | specialty = Hautarzt, city = Berlin |
Pass check: at least five doctors, each with a name and a city.
Results (7 November 2026)
| Wield recipe | Browser agent | |
|---|---|---|
| Runs passing the check | 9 of 9 | 9 of 9 |
| Median time per run | 447 ms | 51.2 s (51,234 ms) |
| Model cost per run | $0 | $0.15 |
| Requests per run (median) | 1 | 152 (to the site: 5) |
| Blocked by the site | 0 runs | 0 of 9 runs |
Wield's fastest run took 66 ms and its slowest took 3.9 s. The recipe had no failed runs.
Why the results differ
The recipe reads the doctor list in one search request per run. It runs from a jameda tab in your own browser, at human pace, so the search uses the session you already have. No model sits in the loop.
The browser agent reached the same list by loading and reading pages. A median run took 6 steps, 152 requests (5 of them to the site), 47,329 input tokens and 2,395 output tokens. Both sides passed every run, so the difference is in speed and cost.
The site did not block the browser agent. The log records no HTTP 401, 403 or 429 errors.
Caveats
- Small sample. Nine runs per side, on three queries.
- Preliminary. The benchmark was still running on 7 November 2026.
- Paced. Runs are paced at about 30 site requests per hour, so counts are low on some sides.
- Model cost only. Wield's $0 covers model cost. It does not count your time.
Terms of jameda
jameda's terms forbid automated retrieval of profile data. The recipe runs one search page at a time that you start, from your own browser. Searches about health are sensitive, so results stay on your device. Check jameda's terms before any automated use.
FAQ
Does jameda block browser agents?
Not in these runs. The browser agent recorded no HTTP 401, 403 or 429 errors from jameda, and it passed all nine runs.
How fast is the Wield recipe for search doctors on jameda?
The median run took 447 ms across nine runs. The fastest took 66 ms and the slowest took 3.9 s. The browser agent median was 51.2 s.
Can I rerun this benchmark?
The method, the task file and the raw logs are described on the benchmark page. The logs record destinations, timing, tokens and checks. They hold no page contents.