transcription results
No champion established.
A champion wins all six dimensions.
- AccuracyNo ordering claim
- CostNo ordering claim
- LatencyNo ordering claim
- ReliabilityNo ordering claim
- EnergyNot measured
- CoverageNot measured
Compare the tools
Accuracy
perfect_transcription_rate · Higher is better
gemini-3-5-transcribe 0.52 0.42–0.62
gemini-2-5-flash 0.57 0.46–0.66
gemini-2-5-pro 0.61 0.51–0.71
gemini-3-5-flash 0.64 0.54–0.74
Cost
USD per unit found · Lower is better
gemini-3-5-transcribe 0.00093
gemini-2-5-flash 0.00090
gemini-2-5-pro 0.0038
gemini-3-5-flash 0.0048
Latency
milliseconds, median · Lower is better
gemini-3-5-transcribe 1500
gemini-2-5-flash 1500
gemini-2-5-pro 3400
gemini-3-5-flash 2300
Reliability
observed error rate · Lower is better
gemini-3-5-transcribe 0.16
gemini-2-5-flash 0
gemini-2-5-pro 0
gemini-3-5-flash 0
Energy
not measured · Lower is better
This run measured no energy draw. The protocol does not freeze one.
Coverage
not measured · Higher is better
This category has no published market census, so coverage is not established.
Check the evidence
Results apply to this test only.
What this does not prove
- Swiss accents and B2B register are not covered by this run (no open corpus available; phase 2).
- 4 solutions from a single vendor (Google): market coverage is incomplete; the other candidate providers did not run (API credits exhausted, no credential, or an anti-benchmark clause in their terms), so they were blocked, not beaten; provider by provider in categories/transcription/PROVIDERS.json and legal/.
- tnorm.v1 does not normalize numbers.
- Wall-clock latency depends on the runner's network.
- Canonical metric = share of clips transcribed perfectly (tnorm.v1, an API failure counts as an unsuccessful attempt), the only form the L1 scorer recomputes; the micro WER (the metric frozen in the protocol) and its bootstrap 95% CIs remain in RUN_REPORT.json, reproducible from events.l2stub.ndjson (amendment A1).
Conflicts and funding
- gemini-3-5-transcribe: devlo is a paying customer of Google, the vendor of this arm, billed per call; devlo is a user of the tool, not a vendor; same conditions and same protocol as any customer, journaled; what this run cost is in FUNDING.json (CONFLICT_REGISTER.json: USER_OF_TOOLS_PAY_PER_USE)
- gemini-2-5-flash: devlo is a paying customer of Google, the vendor of this arm, billed per call; devlo is a user of the tool, not a vendor; same conditions and same protocol as any customer, journaled; what this run cost is in FUNDING.json (CONFLICT_REGISTER.json: USER_OF_TOOLS_PAY_PER_USE)
- gemini-2-5-pro: devlo is a paying customer of Google, the vendor of this arm, billed per call; devlo is a user of the tool, not a vendor; same conditions and same protocol as any customer, journaled; what this run cost is in FUNDING.json (CONFLICT_REGISTER.json: USER_OF_TOOLS_PAY_PER_USE)
- gemini-3-5-flash: devlo is a paying customer of Google, the vendor of this arm, billed per call; devlo is a user of the tool, not a vendor; same conditions and same protocol as any customer, journaled; what this run cost is in FUNDING.json (CONFLICT_REGISTER.json: USER_OF_TOOLS_PAY_PER_USE)
- fonds propres devlo
How to cite
TORNEO, operated by Tempo AI LLC. Run TRANSCRIPTION-001, protocol version 1.0, observed 2 Sep 2026, digest e6377e76.