Model comparison
DeepSeek V4 Pro vs GPT-5.2
DeepSeek V4 Pro and GPT-5.2 score almost the same on the Noometry Index (54.3 vs 54.1), so choose on price, context window or the category you care about most.
Last verified . 41 shared benchmarks.
DeepSeekSummary
- They share 41 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 7 categories and GPT-5.2 in 2 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where GPT-5.2 leads 40.2 to 32.8.
- The biggest single-benchmark swing is ProofBench: 50% for DeepSeek V4 Pro and 15% for GPT-5.2.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 400K.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Pro | GPT-5.2 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 54.3 | 54.1 |
| Released | 2026-04-24 | 2025-12-11 |
| Weights | Open | Proprietary |
| Context window | 1M | 400K |
| Max output | 393K | 128K |
| Input $ / M tokens | $0.66 | $1.75 |
| Output $ / M tokens | $1.98 | $14 |
| Results tracked | 48 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek V4 Pro: 52.4 (#34), GPT-5.2: 51.6 (#37)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| SWE-bench Verified | 77.6% | 73.8% |
| LMArena WebDev | 1582 | 1416 |
| WeirdML | 66.2% | 72.2% |
| LMArena Coding | 1470 | 1447 |
| ALE-Bench | 1,403 | 1,294 |
| FrontierCode | 28.6% | — |
| SWE-bench Verified (bash only) | — | 72.8% |
| SWE-bench Multilingual | — | 66.7% |
| SciCode | 51% | — |
| GSO | — | 27.4% |
| AlgoTune | — | 2.05 |
Agentic & Tool Use GPT-5.2 leads
DeepSeek V4 Pro: 32.8 (#58), GPT-5.2: 40.2 (#24)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| Vending-Bench 2 | 3,285 | 3,591 |
| Terminal-Bench | — | 64.9% |
| APEX-Agents | 47.3% | — |
| Berkeley Function Calling Leaderboard | — | 55.9% |
| GDPval | — | 49.7% |
| Remote Labor Index | — | 2.5% |
| τ²-bench Airline | — | 83% |
| τ²-bench Banking | — | 32.2% |
| τ²-bench Retail | — | 81.6% |
| τ²-bench Telecom | — | 89.7% |
| DeepResearch Bench | — | 41.1% |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), GPT-5.2: 50.2 (#35)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| ARC-AGI-2 | 61.3% | 52.9% |
| Kagi LLM Benchmark | 53.5% | 73.3% |
| NYT Connections (extended) | 91.3% | 83.6% |
| ARC-AGI-1 | 90.5% | 86.2% |
| Chess Puzzles | 47% | 49% |
| LMArena Hard Prompts | 1461 | 1445 |
| Mystery Game Puzzles | 43% | 23% |
| DTBench | 93.9% | 90.9% |
| LMCA | 45.5% | 43.9% |
| Epoch Capabilities Index | 155.31 | 153.45 |
| ForecastBench | 56.1 | 60.1 |
| SimpleBench | — | 45.8% |
| CritPt | 18% | — |
| EnigmaEval | — | 10.4% |
| EBR-Bench | — | 23% |
| Surface Evolver Bench | 40% | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), GPT-5.2: 60.0 (#38)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 64.6% | 67.4% |
| FrontierMath Tier 4 | 26.8% | 31.7% |
| MathArena Final-Answer Competitions | 76.6% | 72% |
| OTIS Mock AIME 2024-2025 | 98.6% | 96.1% |
| ProofBench | 50% | 15% |
| LMArena Math | 1455 | 1440 |
| FrontierMath (Feb 2025 set) | — | 40.7% |
| FrontierMath Tier 4 (v1) | — | 18.8% |
Knowledge Too close to call
DeepSeek V4 Pro: 59.5 (#31), GPT-5.2: 59.3 (#32)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 91.7% | 91.4% |
| SimpleQA Verified | 52.9% | 37.1% |
| Vectara Hallucination Rate | 8.6% | 8.4% |
| LMArena Expert | 1464 | 1445 |
| Humanity's Last Exam | — | 27.8% |
Multimodal Not comparable
DeepSeek V4 Pro: —, GPT-5.2: 51.3 (#7)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| LMArena Vision | — | 1268 |
| VPCT | — | 84% |
| Furniture Assembly | — | 38.3% |
| LMArena Document | — | 1405 |
Multilingual Too close to call
DeepSeek V4 Pro: 54.4 (#45), GPT-5.2: 53.4 (#67)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1439 | 1425 |
| LMArena Chinese | 1486 | 1460 |
| LMArena French | 1472 | 1455 |
| LMArena German | 1458 | 1448 |
| LMArena Japanese | 1445 | 1420 |
| LMArena Korean | 1447 | 1392 |
| LMArena Russian | 1453 | 1440 |
| LMArena Spanish | 1458 | 1433 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), GPT-5.2: 74.7 (#89)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1417 |
Long Context Too close to call
DeepSeek V4 Pro: 45.0 (#51), GPT-5.2: 44.0 (#78)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| LMArena Longer Query | 1458 | 1428 |
| CL-bench | — | 18.2% |
| CL-bench Life | 13.5% | — |
Writing & Preference GPT-5.2 leads
DeepSeek V4 Pro: 65.5 (#46), GPT-5.2: 66.8 (#32)
| Benchmark | DeepSeek V4 Pro | GPT-5.2 |
|---|---|---|
| LMArena Text | 1451 | 1439 |
| LMArena Creative Writing | 1446 | 1401 |
| EQ-Bench Creative Writing | 1553 | 1703 |
| LMArena Multi-Turn | 1467 | 1458 |
| EQ-Bench 4 | 1166 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than GPT-5.2?
DeepSeek V4 Pro and GPT-5.2 score almost the same on the Noometry Index (54.3 vs 54.1), so choose on price, context window or the category you care about most.
Which is cheaper, DeepSeek V4 Pro or GPT-5.2?
DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is DeepSeek V4 Pro or GPT-5.2 better for coding?
They score almost the same on coding (52.4 vs 51.6); test both on your own repository before choosing.
Which has the bigger context window?
DeepSeek V4 Pro does, with 1M tokens against 400K.
How many benchmarks do DeepSeek V4 Pro and GPT-5.2 share?
41 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and GPT-5.2 has 67.