GPT-5.4 mini
OpenAI·
LLMssmallcloud
Overview
Fast reasoning model approaching GPT-5.4 performance at 2x the speed, optimized for sub-agent workloads.
Capabilities and innovations
Reasoning (configurable effort)Multimodal (text + image)Tool use / function callingComputer use2x faster than GPT-5 miniSub-Agent Architecture OptimizedNear-Flagship Performance at Mini Scale2x Speed vs GPT-5 mini
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 88% | unsourced |
| Humanity’s Last Exam (no tools) | 28.2% | unsourced |
| SWE-bench Pro | 54.4% | unsourced |
| MMLU-Pro | 84.55% | unsourced |
| Terminal-Bench 2.x | 60% | unsourced |
| MMMU-Pro | 76.6% | unsourced |
- SWE-bench Verified: benchmark did not exist at release
- MMLU: not reported by the vendor
Links
More from OpenAI
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.