# DeepSeek-R1 vs DeepSeek V4 Flash

> DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 42.0 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-r1-vs-deepseek-v4-flash
- Last updated: 2026-10-10
- Shared benchmarks: 29

## Summary

- They share 29 benchmarks with published results for both. DeepSeek-R1 scores higher in 1 category and DeepSeek V4 Flash in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Flash leads 53.7 to 18.7.
- The biggest single-benchmark swing is ARC-AGI-1: 21.2% for DeepSeek-R1 and 89% for DeepSeek V4 Flash.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.50 / $2.15 for DeepSeek-R1.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 164K.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 42.0 | 53.6 |
| Rank | 110 | 32 |
| Context | 164K | 1M |
| Input $/M | $0.50 | $0.15 |
| Output $/M | $2.15 | $0.60 |
| Weights | Proprietary | Open |

## Coding

- DeepSeek-R1: 45.9 (#63)
- DeepSeek V4 Flash: 47.6 (#56)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| SciCode | 35.7% | 49.9% |
| WeirdML | 41.6% | 63% |
| LMArena Coding | 1427 | 1457 |
| ALE-Bench | 804.12 | 1,306 |
| FrontierCode | — | 18.8% |
| Aider Polyglot | 71.4% | — |
| LMArena WebDev | — | 1581 |
| LiveBench Coding | 66.7% | — |
| AlgoTune | 1.7 | — |

## Agentic & Tool Use

- DeepSeek-R1: 30.4 (#68)
- DeepSeek V4 Flash: —

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
| METR Time Horizons | 53.8% | — |

## Reasoning

- DeepSeek-R1: 18.7 (#253)
- DeepSeek V4 Flash: 53.7 (#25)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| ARC-AGI-2 | 1.3% | 61.4% |
| SimpleBench | 40.8% | 61.1% |
| Kagi LLM Benchmark | 69.4% | 52.2% |
| ARC-AGI-1 | 21.2% | 89% |
| CritPt | 1.1% | 16.6% |
| LMArena Hard Prompts | 1415 | 1443 |
| Epoch Capabilities Index | 141.3 | 154.51 |
| NYT Connections (extended) | — | 89.6% |
| Chess Puzzles | — | 33% |
| LiveBench Reasoning | 83.2% | — |
| Mystery Game Puzzles | — | 34% |
| DTBench | — | 90.9% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 41.7% |
| ForecastBench | 60 | — |
| LiveBench | 71.6% | — |

## Math

- DeepSeek-R1: 42.0 (#70)
- DeepSeek V4 Flash: 60.0 (#31)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 94.4% |
| LMArena Math | 1401 | 1427 |
| FrontierMath (Tiers 1-3) | — | 57.5% |
| FrontierMath Tier 4 | — | 24.4% |
| MathArena Final-Answer Competitions | — | 76.5% |
| ProofBench | — | 56% |
| Omni-MATH | 42.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |

## Knowledge

- DeepSeek-R1: 44.3 (#81)
- DeepSeek V4 Flash: 55.5 (#43)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| GPQA Diamond | 76.3% | 91% |
| LMArena Expert | 1394 | 1441 |
| SimpleQA Verified | — | 33.6% |
| MMLU-Pro | 79.3% | — |
| Confabulations | 12.7% | — |
| Vectara Hallucination Rate | 11.3% | — |
| GPQA (HELM) | 66.6% | — |

## Multilingual

- DeepSeek-R1: 52.6 (#80)
- DeepSeek V4 Flash: 53.2 (#71)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Non-English | 1412 | 1420 |
| LMArena Chinese | 1443 | 1467 |
| LMArena French | 1415 | 1438 |
| LMArena German | 1405 | 1418 |
| LMArena Japanese | 1394 | 1403 |
| LMArena Korean | 1359 | 1383 |
| LMArena Russian | 1423 | 1428 |
| LMArena Spanish | 1410 | 1437 |

## Instruction Following

- DeepSeek-R1: 72.2 (#138)
- DeepSeek V4 Flash: 75.1 (#76)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Instruction Following | 1382 | 1421 |
| LiveBench Instruction Following | 80.5% | — |
| IFEval | 78.4% | — |

## Long Context

- DeepSeek-R1: 45.4 (#36)
- DeepSeek V4 Flash: 43.8 (#81)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Longer Query | 1391 | 1435 |
| Fiction.LiveBench | 75% | — |

## Writing & Preference

- DeepSeek-R1: 61.5 (#84)
- DeepSeek V4 Flash: 64.0 (#59)

| Benchmark | DeepSeek-R1 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Text | 1428 | 1432 |
| LMArena Creative Writing | 1406 | 1403 |
| EQ-Bench Creative Writing | 1500 | 1559 |
| LMArena Multi-Turn | 1404 | 1449 |
| Short-Story Creative Writing | 83% | — |
| WildBench | 82.8% | — |
| LiveBench Language | 48.5% | — |

## FAQ

### Is DeepSeek-R1 better than DeepSeek V4 Flash?

DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 42.0 on the Noometry Index.

### Which is cheaper, DeepSeek-R1 or DeepSeek V4 Flash?

DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; DeepSeek-R1 lists at $0.50 and $2.15.

### Is DeepSeek-R1 or DeepSeek V4 Flash better for coding?

DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 45.9 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Flash does, with 1M tokens against 164K.

### How many benchmarks do DeepSeek-R1 and DeepSeek V4 Flash share?

29 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and DeepSeek V4 Flash has 41.