Compare / GPT-5.4 Nano vs gpt-oss-120b
GPT-5.4 Nano vs gpt-oss-120b
21 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | GPT-5.4 Nano | gpt-oss-120b |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,418 | 1,385 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,405 | 1,351 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,335 | 1,277 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,438 | 1,357 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,387 | 1,323 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,401 | 1,352 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,362 | 1,309 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 56.1 | 30.4 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 20.7 | 11.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 81.7% | 78.2% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 28.3% | 19.6% |
| Artificial Analysisreported | IFBench% of instructions | 75.9% | 69.0% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 76.7% | 52.0% |
| Artificial Analysisreported | SciCode% of problems | 47.2% | 34.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 60.7% | 26.2% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 0.5% | 0.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 42.4% | 23.5% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 27.4% | 12.8% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 76.0% | 65.8% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 100.0% | 99.9% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 3.1% | 14.2% |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.