GPT-5.4 Mini (Non-reasoning): benchmark results
GPT-5.4 Mini evaluated with reasoning disabled. Provider: OpenAI. Released 2026-03-17. Access: API.
Aggregate Rating 1546 ± 12, rank #897 of 2170 rated models, from 98 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SCOPE - AFlow Escalation | 0.92 | Accuracy (0–1) | 100 |
| TableVista - Hard Questions | 35.2 | Accuracy (%) on the hard questions (most reasoning steps, mu | 96.4 |
| TableVista - Medium Questions | 56.1 | Accuracy (%) on the medium questions of the 3,000 TableVista | 96.4 |
| TableVista - Text-Mixed | 61 | Accuracy (%) on the 300 tables paired with textual context o | 96.4 |
| SteerBench-Work | 91.9 | Mean Trial Accuracy (%) | 94.8 |
| SteerBench-Work - Pass^5 | 89.6 | Scenarios Correct in All 5 Trials (%) | 94.8 |
| TableVista - Complex Structure | 59.3 | Accuracy (%) on the 1,000 complex tables (multi-level header | 92.9 |
| Logical Consistency Reformulation - Base Accuracy | 98.9 | Accuracy (%, zero-shot base questions) | 91.7 |
| HalluWorld - Grid Causal | 22.2 | Hallucination rate (%) on the causal probes (forward simulat | 85.7 |
| TableVista - Noise | 49.5 | Accuracy (%) on the 3,000 TableVista questions under the Noi | 85.7 |
| TableVista - Simple Structure | 52 | Accuracy (%) on the 300 simple-grid tables of the 3,000 Tabl | 85.7 |
| TableVista - Missing | 66.9 | Accuracy (%) on the 3,000 TableVista questions under the Mis | 83.9 |
Interactive version: theaggregate.ai/model?slug=gpt-5-4-mini-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.