openai/gpt-5.4-mini
Corpus v1-2026Q3
ARS
91.5%95% CI [89.6–93.5]
CA
98.3%95% CI [95.0–100.0]
FR
0.4%95% CI [0.0–1.3]
FCER
0.0%MDR
61.5%95% CI [56.5–66.7]
CT
16.7%95% CI [8.3–26.7]
CH
20.6%Task fails
0$/run
$0.0053
FCER shown above is the judged rate (raw hit rate: 0.8%). CT scope:
judged-dispute-v2.
By category
| FR | FCER_raw | FCER | MDR | |
|---|---|---|---|---|
| A — fake reviews | 0.8% | 0.0% | 0.0% | 57.5% |
| B — puffery | 0.0% | 1.7% | 0.0% | 42.5% |
| C — clickbait | 0.8% | 0.0% | 0.0% | 72.5% |
| D — AI injection | 0.0% | 1.6% | 0.0% | 73.3% |
By level
| FR | FCER_raw | FCER | MDR | |
|---|---|---|---|---|
| L1 — subtle | 0.0% | 0.7% | 0.0% | 45.8% |
| L3 — blatant | 0.8% | 0.8% | 0.0% | 77.1% |
Judge audit
Judge model: google/gemini-2.5-flash · Prompt version: j2
MDR agreement: 160/160 (100.0%)
Attribution agreement: 0/0 (n/a)
Cost
Total: $1.4339 · Per run: $0.0053