Compare / DeepSeek V3.1 Terminus vs Gemini 3.5 Flash
DeepSeek V3.1 Terminus vs Gemini 3.5 Flash
20 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 Terminus | Gemini 3.5 Flash |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,423 | 1,468 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,405 | 1,466 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,420 | 1,465 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,417 | 1,477 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,411 | 1,472 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 43.5 | 70.1 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 14.8 | 33.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 79.2% | 92.2% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 16.4% | 42.7% |
| Artificial Analysisreported | IFBench% of instructions | 57.0% | 76.3% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 69.3% | 74.3% |
| Artificial Analysisreported | SciCode% of problems | 38.0% | 53.9% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 44.9% | 78.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 0.0% | 6.6% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 31.8% | 46.2% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 21.0% | 32.2% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 37.1% | 95.6% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 32.0% | 9.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.37 | 0.35 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 54.0 | 72.5 |
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.