Compare / Gemini 3.1 Flash Lite Preview vs GPT-5.1
Gemini 3.1 Flash Lite Preview vs GPT-5.1
19 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Gemini 3.1 Flash Lite Preview | GPT-5.1 |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,433 | 1,459 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,421 | 1,448 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,413 | 1,428 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,443 | 1,482 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,407 | 1,450 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,433 | 1,456 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,420 | 1,437 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 34.7 | 49.4 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 15.6 | 24.7 |
| Artificial Analysisreported | GPQA Diamond% of questions | 82.2% | 87.3% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 17.2% | 28.5% |
| Artificial Analysisreported | IFBench% of instructions | 77.2% | 72.9% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 74.3% | 80.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 31.1% | 52.4% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 24.2% | 45.5% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 9.7% | 15.9% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 31.3% | 81.9% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 99.6% | 100.0% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 8.2% | 10.9% |
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.