Model comparison
DeepSeek-R1 vs DeepSeek V4 Pro
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 42.3 on the Noometry Index.
Last verified . 30 shared benchmarks.
DeepSeekSummary
- They share 30 benchmarks with published results for both. DeepSeek-R1 scores higher in 1 category and DeepSeek V4 Pro in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 18.6.
- The biggest single-benchmark swing is ARC-AGI-1: 21.2% for DeepSeek-R1 and 90.5% for DeepSeek V4 Pro.
- DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 164K.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | DeepSeek V4 Pro | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 42.3 | 54.3 |
| Released | 2025-01-20 | 2026-04-24 |
| Weights | Proprietary | Open |
| Context window | 164K | 1M |
| Max output | 64K | 393K |
| Input $ / M tokens | $0.50 | $0.66 |
| Output $ / M tokens | $2.15 | $1.98 |
| Results tracked | 52 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek-R1: 46.3 (#68), DeepSeek V4 Pro: 52.4 (#34)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| SciCode | 35.7% | 51% |
| WeirdML | 41.6% | 66.2% |
| LMArena Coding | 1427 | 1470 |
| ALE-Bench | 804.12 | 1,403 |
| SWE-bench Verified | — | 77.6% |
| FrontierCode | — | 28.6% |
| Aider Polyglot | 71.4% | — |
| LMArena WebDev | — | 1582 |
| LiveBench Coding | 66.7% | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use DeepSeek V4 Pro leads
DeepSeek-R1: 30.7 (#75), DeepSeek V4 Pro: 32.8 (#58)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | — | 47.3% |
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
| METR Time Horizons | 53.8% | — |
| Vending-Bench 2 | — | 3,285 |
Reasoning DeepSeek V4 Pro leads
DeepSeek-R1: 18.6 (#278), DeepSeek V4 Pro: 56.5 (#24)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| ARC-AGI-2 | 1.3% | 61.3% |
| Kagi LLM Benchmark | 69.4% | 53.5% |
| ARC-AGI-1 | 21.2% | 90.5% |
| CritPt | 1.1% | 18% |
| LMArena Hard Prompts | 1416 | 1461 |
| Epoch Capabilities Index | 141.29 | 155.31 |
| ForecastBench | 60 | 56.1 |
| SimpleBench | 40.8% | — |
| NYT Connections (extended) | — | 91.3% |
| Chess Puzzles | — | 47% |
| LiveBench Reasoning | 83.2% | — |
| Mystery Game Puzzles | — | 43% |
| DTBench | — | 93.9% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 45.5% |
| Surface Evolver Bench | — | 40% |
| LiveBench | 71.6% | — |
Math DeepSeek V4 Pro leads
DeepSeek-R1: 43.8 (#79), DeepSeek V4 Pro: 64.8 (#30)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 98.6% |
| LMArena Math | 1400 | 1455 |
| FrontierMath (Tiers 1-3) | — | 64.6% |
| FrontierMath Tier 4 | — | 26.8% |
| MathArena Final-Answer Competitions | — | 76.6% |
| ProofBench | — | 50% |
| Omni-MATH | 42.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
Knowledge DeepSeek V4 Pro leads
DeepSeek-R1: 44.5 (#87), DeepSeek V4 Pro: 59.5 (#31)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 76.3% | 91.7% |
| Vectara Hallucination Rate | 11.3% | 8.6% |
| LMArena Expert | 1394 | 1464 |
| SimpleQA Verified | — | 52.9% |
| MMLU-Pro | 79.3% | — |
| Confabulations | 12.7% | — |
| GPQA (HELM) | 66.6% | — |
Multilingual DeepSeek V4 Pro leads
DeepSeek-R1: 52.4 (#85), DeepSeek V4 Pro: 54.4 (#45)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1412 | 1439 |
| LMArena Chinese | 1442 | 1486 |
| LMArena French | 1417 | 1472 |
| LMArena German | 1404 | 1458 |
| LMArena Japanese | 1391 | 1445 |
| LMArena Korean | 1360 | 1447 |
| LMArena Russian | 1423 | 1453 |
| LMArena Spanish | 1411 | 1458 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek-R1: 72.0 (#143), DeepSeek V4 Pro: 76.1 (#47)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1382 | 1448 |
| LiveBench Instruction Following | 80.5% | — |
| IFEval | 78.4% | — |
Long Context Too close to call
DeepSeek-R1: 45.4 (#36), DeepSeek V4 Pro: 45.0 (#51)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1391 | 1458 |
| Fiction.LiveBench | 75% | — |
| CL-bench Life | — | 13.5% |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek-R1: 61.4 (#88), DeepSeek V4 Pro: 65.5 (#46)
| Benchmark | DeepSeek-R1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1428 | 1451 |
| LMArena Creative Writing | 1405 | 1446 |
| EQ-Bench Creative Writing | 1500 | 1553 |
| LMArena Multi-Turn | 1405 | 1467 |
| Short-Story Creative Writing | 83% | — |
| WildBench | 82.8% | — |
| EQ-Bench 4 | — | 1166 |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than DeepSeek V4 Pro?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 42.3 on the Noometry Index.
Which is cheaper, DeepSeek-R1 or DeepSeek V4 Pro?
DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.
Is DeepSeek-R1 or DeepSeek V4 Pro better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 46.3 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4 Pro does, with 1M tokens against 164K.
How many benchmarks do DeepSeek-R1 and DeepSeek V4 Pro share?
30 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and DeepSeek V4 Pro has 48.