Compare / DeepSeek V4 Pro 0423 vs GPT-5.4
DeepSeek V4 Pro 0423 vs GPT-5.4
28 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-5.4 |
|---|---|---|---|
| Arena (formerly LMArena)reported | Confirmed task successIPS effect estimate | -0.04 | -0.04 |
| Arena (formerly LMArena)reported | Praise over complaintIPS effect estimate | -0.02 | -0.01 |
| Arena (formerly LMArena)reported | SteerabilityIPS effect estimate | -0.03 | 0.02 |
| Arena (formerly LMArena)reported | Tool groundingIPS effect estimate | -0.00 | 0.00 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,455 | 1,485 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,460 | 1,482 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,446 | 1,443 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,480 | 1,515 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,452 | 1,471 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,458 | 1,475 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,451 | 1,465 |
| Microsoft STATE-Benchreported | Consistency (pass^5)% of tasks | 25.3% | 38.0% |
| Microsoft STATE-Benchreported | Customer support task success (pass@1)% of tasks | 45.6% | 57.6% |
| Microsoft STATE-Benchreported | Shopping assistant task success (pass@1)% of tasks | 48.4% | 53.6% |
| Microsoft STATE-Benchreported | Task success (pass@1)% of tasks | 47.2% | 55.7% |
| Microsoft STATE-Benchreported | Travel task success (pass@1)% of tasks | 47.6% | 55.9% |
| Microsoft STATE-Benchreported | User experiencescore from 1 to 5 | 3.36 | 3.49 |
| OpenHands Indexreported | Average% resolved | 40.7% | 64.3% |
| OpenHands Indexreported | Front end (SWE-Bench Multimodal)% resolved | 36.8% | 36.8% |
| OpenHands Indexreported | Greenfield (Commit0)% resolved | 12.5% | 56.2% |
| OpenHands Indexreported | Information gathering (GAIA)% resolved | 12.7% | 82.4% |
| OpenHands Indexreported | Issue resolution (SWE-Bench)% resolved | 73.2% | 75.6% |
| OpenHands Indexreported | Testing (SWT-Bench)% resolved | 68.1% | 70.4% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 34.0% | 30.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.36 | 0.40 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 68.4 | 68.8 |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 97.2% | 99.9% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 8.6% | 7.0% |
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.