GPT-5.4 Mini: benchmark results
OpenAI's smaller, faster GPT-5.4 tier for cost-efficient reasoning workloads. Provider: OpenAI. Released 2026-03-17. Access: API.
Aggregate Rating 1692 ± 3, rank #212 of 1656 rated models, from 1478 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGC-Bench - hypobench | 0.48 | Dataset z-score | 100 |
| ArmBench-LLM - Reading Comprehension | 62.28 | Mean of accuracy and BLEU tasks (0-100) | 100 |
| CLIR-Bench | 50.15 | Macro-average accuracy (%) over the 11 tasks, 4-option multi | 100 |
| CLIR-Bench - Anchor-State Retrieval | 63.33 | Accuracy (%) on anchor-state retrieval (value of one variabl | 100 |
| CLIR-Bench - Cross-Variable Reasoning | 63.33 | Accuracy (%) on asynchronous cross-variable reasoning, 4-opt | 100 |
| CLIR-Bench - Temporal Grounding | 58.33 | Accuracy (%) on temporal grounding (locate the time points o | 100 |
| CLIR-Bench - Time-Series Summarization | 93.33 | Accuracy (%) on clinical time-series summarization, 4-option | 100 |
| CLIR-Bench - Trend Pattern Recognition | 35 | Accuracy (%) on trend pattern recognition within a window, 4 | 100 |
| DexHoldem Agentic Perception - Turn Ownership | 94.4 | Field accuracy (%; whose turn it is, all 36 problems; 36 tab | 100 |
| GACL - Battleship | 85.94 | Normalized tournament points (0-100); source-selected best g | 100 |
| Grounding Awareness - Stateless - Contextual Selectivity | 0.1 | Contextual Selectivity (0–1) | 100 |
| HUSH-Bench (A-Mem) | 8.9 | Unsolicited history integration score (0-100): a GPT-5.4 jud | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-4-mini · How It Works · Data refreshed daily, snapshot 2026-10-20.