Model comparison
DeepSeek-V3.1 vs DeepSeek V4 Pro
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 42.8 on the Noometry Index. DeepSeek-V3.1 costs 2.3× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
DeepSeekSummary
- They share 25 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 0 categories 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 27.9.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 66.2% for DeepSeek V4 Pro.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 164K.
Side by side
| DeepSeek-V3.1 | DeepSeek V4 Pro | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 42.8 | 54.3 |
| Released | 2025-08-21 | 2026-04-24 |
| Weights | Open | Open |
| Context window | 164K | 1M |
| Max output | 8K | 393K |
| Input $ / M tokens | $0.25 | $0.66 |
| Output $ / M tokens | $0.95 | $1.98 |
| Results tracked | 27 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek-V3.1: 40.3 (#144), DeepSeek V4 Pro: 52.4 (#34)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| WeirdML | 38.4% | 66.2% |
| LMArena Coding | 1417 | 1470 |
| SWE-bench Verified | — | 77.6% |
| FrontierCode | — | 28.6% |
| LMArena WebDev | — | 1582 |
| SciCode | — | 51% |
| ALE-Bench | — | 1,403 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, DeepSeek V4 Pro: 32.8 (#58)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | — | 47.3% |
| Vending-Bench 2 | — | 3,285 |
Reasoning DeepSeek V4 Pro leads
DeepSeek-V3.1: 27.9 (#110), DeepSeek V4 Pro: 56.5 (#24)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| Kagi LLM Benchmark | 53.2% | 53.5% |
| LMArena Hard Prompts | 1417 | 1461 |
| DTBench | 82.7% | 93.9% |
| LMCA | 24.3% | 45.5% |
| Epoch Capabilities Index | 139.92 | 155.31 |
| ForecastBench | 58 | 56.1 |
| ARC-AGI-2 | — | 61.3% |
| SimpleBench | 40% | — |
| NYT Connections (extended) | — | 91.3% |
| ARC-AGI-1 | — | 90.5% |
| CritPt | — | 18% |
| Chess Puzzles | — | 47% |
| Mystery Game Puzzles | — | 43% |
| Surface Evolver Bench | — | 40% |
Math DeepSeek V4 Pro leads
DeepSeek-V3.1: 38.9 (#122), DeepSeek V4 Pro: 64.8 (#30)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Math | 1420 | 1455 |
| FrontierMath (Tiers 1-3) | — | 64.6% |
| FrontierMath Tier 4 | — | 26.8% |
| MathArena Final-Answer Competitions | — | 76.6% |
| OTIS Mock AIME 2024-2025 | — | 98.6% |
| ProofBench | — | 50% |
Knowledge DeepSeek V4 Pro leads
DeepSeek-V3.1: 43.7 (#90), DeepSeek V4 Pro: 59.5 (#31)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 8.6% |
| LMArena Expert | 1405 | 1464 |
| GPQA Diamond | — | 91.7% |
| SimpleQA Verified | — | 52.9% |
Multilingual DeepSeek V4 Pro leads
DeepSeek-V3.1: 51.6 (#106), DeepSeek V4 Pro: 54.4 (#45)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1400 | 1439 |
| LMArena Chinese | 1469 | 1486 |
| LMArena French | 1447 | 1472 |
| LMArena German | 1411 | 1458 |
| LMArena Japanese | 1378 | 1445 |
| LMArena Korean | 1337 | 1447 |
| LMArena Russian | 1405 | 1453 |
| LMArena Spanish | 1431 | 1458 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek-V3.1: 73.9 (#110), DeepSeek V4 Pro: 76.1 (#47)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1400 | 1448 |
Long Context DeepSeek V4 Pro leads
DeepSeek-V3.1: 36.3 (#232), DeepSeek V4 Pro: 45.0 (#51)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1422 | 1458 |
| Fiction.LiveBench | 52.8% | — |
| CL-bench Life | — | 13.5% |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek-V3.1: 60.3 (#98), DeepSeek V4 Pro: 65.5 (#46)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1420 | 1451 |
| LMArena Creative Writing | 1401 | 1446 |
| EQ-Bench Creative Writing | 1436 | 1553 |
| LMArena Multi-Turn | 1408 | 1467 |
| EQ-Bench 4 | — | 1166 |
Frequently asked questions
Is DeepSeek-V3.1 better than DeepSeek V4 Pro?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 42.8 on the Noometry Index. DeepSeek-V3.1 costs 2.3× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.1 or DeepSeek V4 Pro?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.
Is DeepSeek-V3.1 or DeepSeek V4 Pro better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 40.3 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-V3.1 and DeepSeek V4 Pro share?
25 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and DeepSeek V4 Pro has 48.