---
run_id: TRANSCRIPTION-001
version_sha256: e6377e76fecdfab6861603a59eec800e0c9cbf27324d59a96396dd79cab0720c
validity_state: UNDER_REVIEW
status_revision: unattested
status_url: https://torneo.ai/api/v1/attestations/TRANSCRIPTION-001.json
as_of: 2026-09-02T18:42:22.745775Z
---

Validity state: Under review

# transcription results, TRANSCRIPTION-001

Do not name a winner: the confidence intervals overlap. Report the range, the n and the observation date.

**Verdict.** No champion established.

**Rule.** A champion wins all six dimensions.

**Next action.** [Explore results](/v8/t/transcription#compare)

- Task: Batch transcription of FR/DE/EN clips of 5 to 30 s, language hint provided, plain-text output
- Metric: Share of recordings transcribed with no error at all
- Population: 90 frozen units
- Participants: gemini-3-5-transcribe (Google), gemini-2-5-flash (Google), gemini-2-5-pro (Google), gemini-3-5-flash (Google)
- Observed: 2026-09-02T18:42:22.745775Z, valid until 2026-12-01T18:42:22Z
- Status: OK
- Version: 1.0
- Result digest (sha256): e6377e76fecdfab6861603a59eec800e0c9cbf27324d59a96396dd79cab0720c
- Protocol lock (sha256): 7056f0e2f8d0d81ca2c9c040b2a065a71256d94e9ec47b0d675a628f898301f6
- Raw data (sha256): 9e870988ea9e0979f0df49f07a1fc10a09f58d13621b67b99b50207cc5611db7
- Evidence: runs/TRANSCRIPTION-001
- Corrections: runs/TRANSCRIPTION-001/ERRATA.md
- Reproduce: `gladiator reproduce runs/TRANSCRIPTION-001`
- Licence: CC-BY-4.0
- Citation: TORNEO, run TRANSCRIPTION-001, observed 2026-09-02T18:42:22.745775Z, status OK: Faithfully transcribe audio into text (speech-to-text via API) measured on FLEURS dev fr_fr/de_de/en_us, read speech, encyclopedic domain; France/Germany/USA accents (90 frozen units), observed 2026-09-02T18:42:22.745775Z. https://torneo.ai/runs/TRANSCRIPTION-001/result.json
- Attribution: TORNEO. CC-BY-4.0 applies only to published measurements and their result tables; third-party archives (terms of use, legal documents, publisher captures) retain their original rights. Licensing does not establish publication or scientific validity.
- Human page: https://torneo.ai/v8/t/transcription

## Dimensions

| Dimension | Unit | Favourable | Status | Because |
|---|---|---|---|---|
| Accuracy | perfect_transcription_rate | higher | No ordering claim | Do not name a winner: the confidence intervals overlap. Report the range, the n and the observation date. |
| Cost | USD per unit found | lower | No ordering claim | Do not name a winner: the confidence intervals overlap. Report the range, the n and the observation date. |
| Latency | milliseconds, median | lower | No ordering claim | Do not name a winner: the confidence intervals overlap. Report the range, the n and the observation date. |
| Reliability | observed error rate | lower | No ordering claim | Do not name a winner: the confidence intervals overlap. Report the range, the n and the observation date. |
| Energy | not measured | lower | Not measured | This run measured no energy draw. The protocol does not freeze one. |
| Coverage | not measured | higher | Not measured | This category has no published market census, so coverage is not established. |

### Accuracy

Unit: perfect_transcription_rate. Higher is better.

| Participant | Value | 95% interval | Status |
|---|---|---|---|
| gemini-3-5-transcribe | 0.52 | 0.42 to 0.62 | OK |
| gemini-2-5-flash | 0.57 | 0.46 to 0.66 | OK |
| gemini-2-5-pro | 0.61 | 0.51 to 0.71 | OK |
| gemini-3-5-flash | 0.64 | 0.54 to 0.74 | OK |

### Cost

Unit: USD per unit found. Lower is better.

| Participant | Value | 95% interval | Status |
|---|---|---|---|
| gemini-3-5-transcribe | 0.00093 | not published | OK |
| gemini-2-5-flash | 0.00090 | not published | OK |
| gemini-2-5-pro | 0.0038 | not published | OK |
| gemini-3-5-flash | 0.0048 | not published | OK |

### Latency

Unit: milliseconds, median. Lower is better.

| Participant | Value | 95% interval | Status |
|---|---|---|---|
| gemini-3-5-transcribe | 1500 | not published | OK |
| gemini-2-5-flash | 1500 | not published | OK |
| gemini-2-5-pro | 3400 | not published | OK |
| gemini-3-5-flash | 2300 | not published | OK |

### Reliability

Unit: observed error rate. Lower is better.

| Participant | Value | 95% interval | Status |
|---|---|---|---|
| gemini-3-5-transcribe | 0.16 | not published | OK |
| gemini-2-5-flash | 0 | not published | OK |
| gemini-2-5-pro | 0 | not published | OK |
| gemini-3-5-flash | 0 | not published | OK |

## Limits

- 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 of interest

- 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)

Funding: fonds propres devlo.

## How to cite

TORNEO, operated by Tempo AI LLC. Cite the run id, the observation date and the result digest.