Compare / Phi 4 vs GPT-5 Mini
Phi 4 vs GPT-5 Mini
16 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Phi 4 | GPT-5 Mini |
|---|---|---|---|
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,256 | 1,381 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,210 | 1,324 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,268 | 1,400 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,245 | 1,373 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,256 | 1,390 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,229 | 1,352 |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Irrelevance detection% correct | 87.5% | 91.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Memory% correct | 24.7% | 44.3% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Multi-turn tasks% correct | 3.9% | 27.5% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Overall accuracy% correct | 28.8% | 55.5% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Relevance detection% correct | 50.0% | 93.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (curated)% correct | 69.6% | 69.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (user-contributed)% correct | 60.7% | 62.5% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Web search% correct | 4.5% | 82.0% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 80.7% | 99.9% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 3.7% | 12.9% |
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.