OPENAI · MODEL TELEMETRY CARD
GPT-5.4
Canonical UID: openai/gpt-5.4-20260305 · Released: 2026-03-05
| Domain | Benchmark | Effort / Mode | ⚡ Value/Cost # | 🏆 Performance # | Score | Task $ | Unit Cost $ | Unit |
|---|---|---|---|---|---|---|---|---|
| Code | ALE-Bench | STD | #43 / 94 | #47 / 94 | 1086.03 | $0.052314 | $0.00004817 | rating point |
| Code | ALE-Bench | MEDIUM | #6 / 94 | #20 / 94 | 1520.72 | $0.26258 | $0.0001727 | rating point |
| Code | ALE-Bench | HIGH | #12 / 94 | #15 / 94 | 1607 | $0.612174 | $0.0003809 | rating point |
| Reason | ProofBench | XHIGH | #29 / 62 | #23 / 62 | 0.56 | $3.20201 | $5.718 | correct answer |
| Agent | DeepSWE | XHIGH | #43 / 65 | #45 / 65 | 0.518 | $5.65246 | $10.92 | solved task |
Compare against all models on ← #frontier Live Ranking Board · Fixed-capability cost half-life analysis on /methodology.