GPT-5.4 nano
OpenAI·
LLMssmallcloud
Overview
Smallest, cheapest GPT-5.4 variant for classification, extraction, and cost-sensitive sub-agent tasks.
Capabilities and innovations
Multimodal (text + image)Tool use / function callingClassification & extractionCost-optimized ($0.20/1M input)Ultra Cost-Efficient ReasoningSub-Agent Era DesignSmallest GPT-5.4 Variant
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| GPQA Diamond | 82.8% | unsourced |
| Humanity’s Last Exam (no tools) | 24.3% | unsourced |
| SWE-bench Pro | 52.4% | unsourced |
| Terminal-Bench 2.x | 41.57% | independent Terminal-Bench 2.1 41.57%, measured by Vals AI with the Terminus 2 harness (accessed 2026-09-23). Vals AI is the reference harness for this field so values compare across vendors; vendor and other third-party figures stay in the notes below. Replaces 46.3 (unsourced). |
| MMMU-Pro | 66.1% | unsourced |
- SWE-bench Verified: benchmark did not exist at release
- MMLU: not reported by the vendor
- MMLU-Pro: 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.