Compare / DeepSeek V3.1 vs GPT-5.4 Nano
DeepSeek V3.1 vs GPT-5.4 Nano
15 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | DeepSeek V3.1 | GPT-5.4 Nano |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,417 | 1,405 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,402 | 1,335 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,432 | 1,438 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,416 | 1,387 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,417 | 1,401 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,407 | 1,362 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 13.7 | 20.7 |
| Artificial Analysisreported | GPQA Diamond% of questions | 77.9% | 81.7% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 14.3% | 28.3% |
| Artificial Analysisreported | IFBench% of instructions | 41.5% | 75.9% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 56.7% | 76.7% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 25.0% | 42.4% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 37.4% | 76.0% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 94.5% | 100.0% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 5.5% | 3.1% |
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.