Compare / DeepSeek V4 Pro 0423 vs GPT-5 Mini
DeepSeek V4 Pro 0423 vs GPT-5 Mini
20 results from 5 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | DeepSeek V4 Pro 0423 | GPT-5 Mini |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,455 | 1,394 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,460 | 1,381 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,446 | 1,324 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,480 | 1,400 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,452 | 1,373 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,458 | 1,390 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,451 | 1,352 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 68.8 | 15.6 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 36.0 | 20.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.8% | 82.8% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 41.0% | 21.5% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 80.3% | 72.3% |
| Artificial Analysisreported | SciCode% of problems | 51.0% | 39.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 78.7% | 3.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 14.1% | 0.0% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 39.6% | 15.5% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 47.0% | 21.6% |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 97.2% | 99.9% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 8.6% | 12.9% |
| Vending-Bench 2reported | Money after a yearUS dollars | $3,284.52 | $-31.18 |
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.