Model comparison
DeepSeek-R1 vs GPT-4.1
DeepSeek-R1 is the stronger model overall, scoring 42.0 to 35.7 on the Noometry Index.
Last verified . 37 shared benchmarks.
DeepSeekSummary
- They share 37 benchmarks with published results for both. DeepSeek-R1 scores higher in 8 categories and GPT-4.1 in 1 category; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-R1 leads 42.0 to 21.4.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 66.4% for DeepSeek-R1 and 38.3% for GPT-4.1.
- DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $2 / $8 for GPT-4.1.
- GPT-4.1 accepts more context: 1.05M tokens versus 164K.
Side by side
| DeepSeek-R1 | GPT-4.1 | |
|---|---|---|
| Provider | DeepSeek | OpenAI |
| Noometry Index | 42.0 | 35.7 |
| Released | 2025-01-20 | 2025-04-14 |
| Weights | Proprietary | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 64K | 33K |
| Input $ / M tokens | $0.50 | $2 |
| Output $ / M tokens | $2.15 | $8 |
| Results tracked | 52 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1 leads
DeepSeek-R1: 45.9 (#63), GPT-4.1: 34.2 (#228)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| Aider Polyglot | 71.4% | 52.4% |
| WeirdML | 41.6% | 39% |
| LMArena Coding | 1427 | 1391 |
| ALE-Bench | 804.12 | 558.1 |
| SWE-bench Verified | — | 48.5% |
| SWE-bench Verified (bash only) | — | 39.6% |
| SciCode | 35.7% | — |
| LiveBench Coding | 66.7% | — |
| CadEval | — | 42% |
| AlgoTune | 1.7 | — |
Agentic & Tool Use GPT-4.1 leads
DeepSeek-R1: 30.4 (#68), GPT-4.1: 34.4 (#40)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 54% |
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
| METR Time Horizons | 53.8% | — |
Reasoning DeepSeek-R1 leads
DeepSeek-R1: 18.7 (#253), GPT-4.1: 11.7 (#334)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| ARC-AGI-2 | 1.3% | 0.4% |
| SimpleBench | 40.8% | 27% |
| Kagi LLM Benchmark | 69.4% | 52.3% |
| ARC-AGI-1 | 21.2% | 5.5% |
| LMArena Hard Prompts | 1415 | 1384 |
| Epoch Capabilities Index | 141.3 | 136.79 |
| ForecastBench | 60 | 61.5 |
| CritPt | 1.1% | — |
| Chess Puzzles | — | 6% |
| EnigmaEval | — | 2.2% |
| LiveBench Reasoning | 83.2% | — |
| DTBench | — | 68.3% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 25.6% |
| LiveBench | 71.6% | — |
Math DeepSeek-R1 leads
DeepSeek-R1: 42.0 (#70), GPT-4.1: 21.4 (#269)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 38.3% |
| Omni-MATH | 42.4% | 47.1% |
| LMArena Math | 1401 | 1369 |
| MATH Level 5 | 96.6% | 83% |
| FrontierMath (Tiers 1-3) | — | 6% |
| LiveBench Math | 80.7% | — |
| FrontierMath (Feb 2025 set) | — | 5.5% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge DeepSeek-R1 leads
DeepSeek-R1: 44.3 (#81), GPT-4.1: 37.0 (#152)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| GPQA Diamond | 76.3% | 66.9% |
| MMLU-Pro | 79.3% | 81.1% |
| Vectara Hallucination Rate | 11.3% | 5.6% |
| GPQA (HELM) | 66.6% | 65.9% |
| LMArena Expert | 1394 | 1364 |
| Humanity's Last Exam | — | 5.4% |
| SimpleQA Verified | — | 31.1% |
| Confabulations | 12.7% | — |
Multimodal Not comparable
DeepSeek-R1: —, GPT-4.1: 37.5 (#66)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| LMArena Vision | — | 1210 |
| GeoBench | — | 72% |
Multilingual DeepSeek-R1 leads
DeepSeek-R1: 52.6 (#80), GPT-4.1: 49.6 (#128)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| LMArena Non-English | 1412 | 1370 |
| LMArena Chinese | 1443 | 1382 |
| LMArena French | 1415 | 1383 |
| LMArena German | 1405 | 1380 |
| LMArena Japanese | 1394 | 1320 |
| LMArena Korean | 1359 | 1339 |
| LMArena Russian | 1423 | 1377 |
| LMArena Spanish | 1410 | 1375 |
Instruction Following Too close to call
DeepSeek-R1: 72.2 (#138), GPT-4.1: 71.5 (#148)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| IFEval | 78.4% | 83.8% |
| LMArena Instruction Following | 1382 | 1367 |
| LiveBench Instruction Following | 80.5% | — |
Long Context DeepSeek-R1 leads
DeepSeek-R1: 45.4 (#36), GPT-4.1: 40.0 (#158)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| Fiction.LiveBench | 75% | 63.9% |
| LMArena Longer Query | 1391 | 1385 |
Writing & Preference DeepSeek-R1 leads
DeepSeek-R1: 61.5 (#84), GPT-4.1: 57.8 (#120)
| Benchmark | DeepSeek-R1 | GPT-4.1 |
|---|---|---|
| LMArena Text | 1428 | 1383 |
| LMArena Creative Writing | 1406 | 1363 |
| EQ-Bench Creative Writing | 1500 | 1420 |
| WildBench | 82.8% | 85.4% |
| LMArena Multi-Turn | 1404 | 1398 |
| Short-Story Creative Writing | 83% | — |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than GPT-4.1?
DeepSeek-R1 is the stronger model overall, scoring 42.0 to 35.7 on the Noometry Index.
Which is cheaper, DeepSeek-R1 or GPT-4.1?
DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; GPT-4.1 lists at $2 and $8.
Is DeepSeek-R1 or GPT-4.1 better for coding?
DeepSeek-R1 scores higher on coding benchmarks: 45.9 versus 34.2 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-R1 and GPT-4.1 share?
37 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and GPT-4.1 has 52.