Compare / Gemma 4 26B A4B vs GPT-5.6 Terra
Gemma 4 26B A4B vs GPT-5.6 Terra
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 | Gemma 4 26B A4B | GPT-5.6 Terra |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,445 | 1,467 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,444 | 1,466 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,399 | 1,428 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,475 | 1,504 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,437 | 1,462 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,438 | 1,465 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,406 | 1,445 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 39.3 | 76.7 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 16.7 | 42.1 |
| Artificial Analysisreported | GPQA Diamond% of questions | 79.2% | 92.5% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 19.3% | 42.9% |
| Artificial Analysisreported | IFBench% of instructions | 72.4% | 71.2% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 65.7% | 83.0% |
| Artificial Analysisreported | SciCode% of problems | 40.0% | 55.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 39.0% | 88.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 25.0% | 62.9% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 12.0% | 40.2% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 43.6% | 86.3% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 8.0% | 22.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.35 | 0.39 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 43.8 | 60.4 |
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.