Model comparison
DeepSeek V4 Pro vs GPT-4.5
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 37.2 on the Noometry Index.
Last verified . 24 shared benchmarks.
DeepSeekSummary
- They share 24 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and GPT-4.5 in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 13.9.
- The biggest single-benchmark swing is ARC-AGI-1: 90.5% for DeepSeek V4 Pro and 10.3% for GPT-4.5.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Pro | GPT-4.5 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 54.3 | 37.2 |
| Released | 2026-04-24 | 2025-02-27 |
| Weights | Open | Proprietary |
| Context window | 1M | — |
| Max output | 393K | — |
| Input $ / M tokens | $0.66 | — |
| Output $ / M tokens | $1.98 | — |
| Results tracked | 48 | 42 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), GPT-4.5: 42.2 (#109)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| WeirdML | 66.2% | 39.4% |
| LMArena Coding | 1470 | 1396 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| Aider Polyglot | — | 44.9% |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| LiveBench Coding | — | 75.2% |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use DeepSeek V4 Pro leads
DeepSeek V4 Pro: 32.8 (#58), GPT-4.5: 27.9 (#97)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| APEX-Agents | 47.3% | — |
| Cybench | — | 17.5% |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), GPT-4.5: 13.9 (#330)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| ARC-AGI-2 | 61.3% | 0.8% |
| ARC-AGI-1 | 90.5% | 10.3% |
| LMArena Hard Prompts | 1461 | 1403 |
| Epoch Capabilities Index | 155.31 | 136.74 |
| ForecastBench | 56.1 | 61.7 |
| SimpleBench | — | 34.5% |
| Kagi LLM Benchmark | 53.5% | — |
| NYT Connections (extended) | 91.3% | — |
| CritPt | 18% | — |
| Chess Puzzles | 47% | — |
| EnigmaEval | — | 3.2% |
| LiveBench Reasoning | — | 71.1% |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LiveBench Data Analysis | — | 64.3% |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| LiveBench | — | 69% |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), GPT-4.5: 32.6 (#211)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 37.8% |
| LMArena Math | 1455 | 1412 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| LiveBench Math | — | 69.3% |
| MATH Level 5 | — | 78.6% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), GPT-4.5: 32.5 (#211)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| GPQA Diamond | 91.7% | 68.7% |
| LMArena Expert | 1464 | 1394 |
| Humanity's Last Exam | — | 5.4% |
| SimpleQA Verified | 52.9% | — |
| Confabulations | — | 13.6% |
| Vectara Hallucination Rate | 8.6% | — |
Multimodal Not comparable
DeepSeek V4 Pro: —, GPT-4.5: 37.6 (#71)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| LMArena Vision | — | 1195 |
| VPCT | — | 45% |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), GPT-4.5: 52.5 (#83)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| LMArena Non-English | 1439 | 1413 |
| LMArena Chinese | 1486 | 1421 |
| LMArena French | 1472 | 1418 |
| LMArena German | 1458 | 1457 |
| LMArena Japanese | 1445 | 1416 |
| LMArena Korean | 1447 | 1392 |
| LMArena Russian | 1453 | 1419 |
| LMArena Spanish | 1458 | — |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), GPT-4.5: 72.6 (#134)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1404 |
| LiveBench Instruction Following | — | 72.3% |
Long Context DeepSeek V4 Pro leads
DeepSeek V4 Pro: 45.0 (#51), GPT-4.5: 40.4 (#155)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| LMArena Longer Query | 1458 | 1406 |
| Fiction.LiveBench | — | 63.9% |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), GPT-4.5: 56.9 (#134)
| Benchmark | DeepSeek V4 Pro | GPT-4.5 |
|---|---|---|
| LMArena Text | 1451 | 1417 |
| LMArena Creative Writing | 1446 | 1394 |
| EQ-Bench Creative Writing | 1553 | 1258 |
| LMArena Multi-Turn | 1467 | 1444 |
| Short-Story Creative Writing | — | 75.6% |
| EQ-Bench 4 | 1166 | — |
| LiveBench Language | — | 61.5% |
Frequently asked questions
Is DeepSeek V4 Pro better than GPT-4.5?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 37.2 on the Noometry Index.
Is DeepSeek V4 Pro or GPT-4.5 better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 42.2 in the Noometry coding category.
How many benchmarks do DeepSeek V4 Pro and GPT-4.5 share?
24 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GPT-4.5 has 42.