Model comparison
DeepSeek V4 Flash vs DeepSeek V4 Pro
DeepSeek V4 Flash and DeepSeek V4 Pro score almost the same on the Noometry Index (53.6 vs 54.3), so choose on price, context window or the category you care about most.
Last verified . 40 shared benchmarks.
DeepSeekSummary
- They share 40 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 0 categories and DeepSeek V4 Pro in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Pro leads 64.8 to 60.3.
- The biggest single-benchmark swing is SimpleQA Verified: 33.6% for DeepSeek V4 Flash and 52.9% for DeepSeek V4 Pro.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
Side by side
| DeepSeek V4 Flash | DeepSeek V4 Pro | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 53.6 | 54.3 |
| Released | 2026-04-24 | 2026-04-24 |
| Weights | Open | Open |
| Context window | 1M | 1M |
| Max output | 393K | 393K |
| Input $ / M tokens | $0.15 | $0.66 |
| Output $ / M tokens | $0.60 | $1.98 |
| Results tracked | 41 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Flash: 47.9 (#59), DeepSeek V4 Pro: 52.4 (#34)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| FrontierCode | 18.8% | 28.6% |
| LMArena WebDev | 1582 | 1582 |
| SciCode | 49.9% | 51% |
| WeirdML | 63% | 66.2% |
| LMArena Coding | 1457 | 1470 |
| ALE-Bench | 1,306 | 1,403 |
| SWE-bench Verified | — | 77.6% |
Agentic & Tool Use Not comparable
DeepSeek V4 Flash: —, DeepSeek V4 Pro: 32.8 (#58)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | — | 47.3% |
| Vending-Bench 2 | — | 3,285 |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Flash: 53.7 (#30), DeepSeek V4 Pro: 56.5 (#24)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| ARC-AGI-2 | 61.4% | 61.3% |
| Kagi LLM Benchmark | 52.2% | 53.5% |
| NYT Connections (extended) | 89.6% | 91.3% |
| ARC-AGI-1 | 89% | 90.5% |
| CritPt | 16.6% | 18% |
| Chess Puzzles | 33% | 47% |
| LMArena Hard Prompts | 1444 | 1461 |
| Mystery Game Puzzles | 34% | 43% |
| DTBench | 90.9% | 93.9% |
| LMCA | 41.7% | 45.5% |
| Epoch Capabilities Index | 154.49 | 155.31 |
| SimpleBench | 61.1% | — |
| Surface Evolver Bench | — | 40% |
| ForecastBench | — | 56.1 |
Math DeepSeek V4 Pro leads
DeepSeek V4 Flash: 60.3 (#37), DeepSeek V4 Pro: 64.8 (#30)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| FrontierMath (Tiers 1-3) | 57.5% | 64.6% |
| FrontierMath Tier 4 | 24.4% | 26.8% |
| MathArena Final-Answer Competitions | 76.5% | 76.6% |
| OTIS Mock AIME 2024-2025 | 94.4% | 98.6% |
| ProofBench | 56% | 50% |
| LMArena Math | 1427 | 1455 |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Flash: 55.4 (#48), DeepSeek V4 Pro: 59.5 (#31)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 91% | 91.7% |
| SimpleQA Verified | 33.6% | 52.9% |
| LMArena Expert | 1441 | 1464 |
| Vectara Hallucination Rate | — | 8.6% |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Flash: 53.0 (#72), DeepSeek V4 Pro: 54.4 (#45)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1420 | 1439 |
| LMArena Chinese | 1468 | 1486 |
| LMArena French | 1439 | 1472 |
| LMArena German | 1418 | 1458 |
| LMArena Japanese | 1406 | 1445 |
| LMArena Korean | 1384 | 1447 |
| LMArena Russian | 1428 | 1453 |
| LMArena Spanish | 1436 | 1458 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Flash: 74.9 (#81), DeepSeek V4 Pro: 76.1 (#47)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1421 | 1448 |
Long Context DeepSeek V4 Pro leads
DeepSeek V4 Flash: 43.8 (#85), DeepSeek V4 Pro: 45.0 (#51)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1434 | 1458 |
| CL-bench Life | — | 13.5% |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Flash: 63.8 (#61), DeepSeek V4 Pro: 65.5 (#46)
| Benchmark | DeepSeek V4 Flash | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1432 | 1451 |
| LMArena Creative Writing | 1403 | 1446 |
| EQ-Bench Creative Writing | 1559 | 1553 |
| LMArena Multi-Turn | 1449 | 1467 |
| EQ-Bench 4 | — | 1166 |
Frequently asked questions
Is DeepSeek V4 Flash better than DeepSeek V4 Pro?
DeepSeek V4 Flash and DeepSeek V4 Pro score almost the same on the Noometry Index (53.6 vs 54.3), so choose on price, context window or the category you care about most.
Which is cheaper, DeepSeek V4 Flash or DeepSeek V4 Pro?
DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.
Is DeepSeek V4 Flash or DeepSeek V4 Pro better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 47.9 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do DeepSeek V4 Flash and DeepSeek V4 Pro share?
40 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and DeepSeek V4 Pro has 48.