Model comparison
DeepSeek-V3.1 vs DeepSeek V4.1 Flash
DeepSeek V4.1 Flash is the stronger model overall, scoring 52.8 to 42.8 on the Noometry Index.
Last verified . 21 shared benchmarks.
DeepSeekSummary
- They share 21 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 0 categories and DeepSeek V4.1 Flash in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4.1 Flash leads 66.7 to 38.9.
- The biggest single-benchmark swing is LMCA: 24.3% for DeepSeek-V3.1 and 47% for DeepSeek V4.1 Flash.
- DeepSeek V4.1 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.25 / $0.95 for DeepSeek-V3.1.
- DeepSeek V4.1 Flash accepts more context: 1M tokens versus 164K.
Side by side
| DeepSeek-V3.1 | DeepSeek V4.1 Flash | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 42.8 | 52.8 |
| Released | 2025-08-21 | 2026-09-09 |
| Weights | Open | Open |
| Context window | 164K | 1M |
| Max output | 8K | 393K |
| Input $ / M tokens | $0.25 | $0.15 |
| Output $ / M tokens | $0.95 | $0.60 |
| Results tracked | 27 | 37 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 40.3 (#144), DeepSeek V4.1 Flash: 52.9 (#32)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Coding | 1417 | 1506 |
| LMArena WebDev | — | 1619 |
| SciCode | — | 51.9% |
| WeirdML | 38.4% | — |
| ALE-Bench | — | 1,092 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, DeepSeek V4.1 Flash: 31.2 (#69)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| APEX-Agents | — | 39.5% |
| GDP.pdf | — | 19.8% |
Reasoning DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 27.9 (#110), DeepSeek V4.1 Flash: 50.2 (#36)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Hard Prompts | 1417 | 1483 |
| DTBench | 82.7% | 89.9% |
| LMCA | 24.3% | 47% |
| Epoch Capabilities Index | 139.92 | 154.9 |
| SimpleBench | 40% | — |
| Kagi LLM Benchmark | 53.2% | — |
| NYT Connections (extended) | — | 89.6% |
| CritPt | — | 14.3% |
| Mystery Game Puzzles | — | 43% |
| Surface Evolver Bench | — | 46.3% |
| ForecastBench | 58 | — |
Math DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 38.9 (#122), DeepSeek V4.1 Flash: 66.7 (#25)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Math | 1420 | 1477 |
| FrontierMath (Tiers 1-3) | — | 67.4% |
| FrontierMath Tier 4 | — | 26.8% |
| OTIS Mock AIME 2024-2025 | — | 98.3% |
| ProofBench | — | 54% |
Knowledge DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 43.7 (#90), DeepSeek V4.1 Flash: 57.9 (#38)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Expert | 1405 | 1506 |
| GPQA Diamond | — | 89.8% |
| Vectara Hallucination Rate | 5.5% | — |
Multimodal Not comparable
DeepSeek-V3.1: —, DeepSeek V4.1 Flash: 39.1 (#61)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Vision | — | 1277 |
| Furniture Assembly | — | 34.2% |
Multilingual DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 51.6 (#106), DeepSeek V4.1 Flash: 55.0 (#35)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Non-English | 1400 | 1448 |
| LMArena Chinese | 1469 | 1497 |
| LMArena French | 1447 | 1452 |
| LMArena German | 1411 | 1484 |
| LMArena Japanese | 1378 | 1412 |
| LMArena Korean | 1337 | 1452 |
| LMArena Russian | 1405 | 1471 |
| LMArena Spanish | 1431 | 1459 |
Instruction Following DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 73.9 (#110), DeepSeek V4.1 Flash: 77.3 (#26)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Instruction Following | 1400 | 1474 |
Long Context DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 36.3 (#232), DeepSeek V4.1 Flash: 45.2 (#47)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Longer Query | 1422 | 1475 |
| Fiction.LiveBench | 52.8% | — |
Writing & Preference DeepSeek V4.1 Flash leads
DeepSeek-V3.1: 60.3 (#98), DeepSeek V4.1 Flash: 65.4 (#48)
| Benchmark | DeepSeek-V3.1 | DeepSeek V4.1 Flash |
|---|---|---|
| LMArena Text | 1420 | 1462 |
| LMArena Creative Writing | 1401 | 1435 |
| EQ-Bench Creative Writing | 1436 | 1540 |
| LMArena Multi-Turn | 1408 | 1457 |
Frequently asked questions
Is DeepSeek-V3.1 better than DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash is the stronger model overall, scoring 52.8 to 42.8 on the Noometry Index.
Which is cheaper, DeepSeek-V3.1 or DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; DeepSeek-V3.1 lists at $0.25 and $0.95.
Is DeepSeek-V3.1 or DeepSeek V4.1 Flash better for coding?
DeepSeek V4.1 Flash scores higher on coding benchmarks: 52.9 versus 40.3 in the Noometry coding category.
Which has the bigger context window?
DeepSeek V4.1 Flash does, with 1M tokens against 164K.
How many benchmarks do DeepSeek-V3.1 and DeepSeek V4.1 Flash share?
21 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and DeepSeek V4.1 Flash has 37.