Which AI tool really works?
Same task. Same rules. Every result dated.
See the results →Follow the runs
done · 2 published categories
Replay of a recorded resulttranscription ·
- gemini-3-5-transcribe0.520.42 – to 0.62
- gemini-2-5-flash0.570.46 – to 0.66
- gemini-2-5-pro0.610.51 – to 0.71
- gemini-3-5-flash0.640.54 – to 0.74
0.30.40.50.60.70.8
where intervals overlap
Measured with intervals across 2 replicated blocks; the intervals overlap, no tool is ranked first.
Exact intervals
This is our experience, on our corpus, at this date. It is not buying advice.
gemini-3-5-transcribegemini-2-5-flashgemini-2-5-progemini-3-5-flashdropcontactlushagemini-3-5-transcribegemini-2-5-flashgemini-2-5-progemini-3-5-flashdropcontactlushagemini-3-5-transcribegemini-2-5-flashgemini-2-5-progemini-3-5-flashdropcontactlushagemini-3-5-transcribegemini-2-5-flashgemini-2-5-progemini-3-5-flashdropcontactlusha
Results
transcription
→Measured, no tool first
- gemini-3-5-transcribe
- gemini-2-5-flash
- gemini-2-5-pro
- gemini-3-5-flash
emailfind
→No rank demonstrated
- dropcontact
- lusha
Beta runs
partial · Played measurements; publication checks remain separate.
How it works
- Lock the rules
- Run every tool
- Publish with the date
Need a tool measured?
Request a tournament →Operated by Tempo AI LLC, accountable for the calculation, publication and correction of every figure published here. No vendor pays to be measured; TORNEO buys what it tests.