# DeepSeek V4 Flash vs Llama 4 Maverick

> DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 30.6 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-flash-vs-llama-4-maverick
- Last updated: 2026-10-10
- Shared benchmarks: 32

## Summary

- They share 32 benchmarks with published results for both. DeepSeek V4 Flash scores higher in 8 categories and Llama 4 Maverick in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Flash leads 53.7 to 10.1.
- The biggest single-benchmark swing is ARC-AGI-1: 89% for DeepSeek V4 Flash and 4.4% for Llama 4 Maverick.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $0.19 / $0.65 for Llama 4 Maverick.
- DeepSeek V4 Flash accepts more context: 1M tokens versus 128K.

## Snapshot

| | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| Provider | DeepSeek | Meta |
| Noometry Index | 53.6 | 30.6 |
| Rank | 32 | 276 |
| Context | 1M | 128K |
| Input $/M | $0.15 | $0.19 |
| Output $/M | $0.60 | $0.65 |
| Weights | Open | Open |

## Coding

- DeepSeek V4 Flash: 47.6 (#56)
- Llama 4 Maverick: 26.3 (#318)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| SciCode | 49.9% | 33.1% |
| WeirdML | 63% | 24.5% |
| LMArena Coding | 1457 | 1302 |
| ALE-Bench | 1,306 | 172.97 |
| FrontierCode | 18.8% | — |
| SWE-bench Verified (bash only) | — | 21% |
| Aider Polyglot | — | 15.6% |
| LMArena WebDev | 1581 | — |
| BigCodeBench Instruct | — | 49.7% |
| BigCodeBench Complete | — | 61.4% |

## Agentic & Tool Use

- DeepSeek V4 Flash: —
- Llama 4 Maverick: 27.9 (#86)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 37.3% |

## Reasoning

- DeepSeek V4 Flash: 53.7 (#25)
- Llama 4 Maverick: 10.1 (#337)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| ARC-AGI-2 | 61.4% | 0% |
| SimpleBench | 61.1% | 27.7% |
| Kagi LLM Benchmark | 52.2% | 55.9% |
| NYT Connections (extended) | 89.6% | 8% |
| ARC-AGI-1 | 89% | 4.4% |
| CritPt | 16.6% | 0% |
| LMArena Hard Prompts | 1443 | 1281 |
| DTBench | 90.9% | 61.9% |
| LMCA | 41.7% | 15.9% |
| Epoch Capabilities Index | 154.51 | 132.2 |
| Chess Puzzles | 33% | — |
| EnigmaEval | — | 0.6% |
| Mystery Game Puzzles | 34% | — |
| ForecastBench | — | 57.5 |

## Math

- DeepSeek V4 Flash: 60.0 (#31)
- Llama 4 Maverick: 24.6 (#256)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 94.4% | 20.6% |
| LMArena Math | 1427 | 1299 |
| FrontierMath (Tiers 1-3) | 57.5% | — |
| FrontierMath Tier 4 | 24.4% | — |
| MathArena Final-Answer Competitions | 76.5% | — |
| ProofBench | 56% | — |
| Omni-MATH | — | 42.2% |
| MATH Level 5 | — | 73% |
| FrontierMath (Feb 2025 set) | — | 0.7% |

## Knowledge

- DeepSeek V4 Flash: 55.5 (#43)
- Llama 4 Maverick: 33.3 (#196)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| GPQA Diamond | 91% | 67% |
| LMArena Expert | 1441 | 1259 |
| Humanity's Last Exam | — | 5.7% |
| SimpleQA Verified | 33.6% | — |
| MMLU-Pro | — | 81% |
| Confabulations | — | 22.6% |
| Vectara Hallucination Rate | — | 8.2% |
| GPQA (HELM) | — | 65% |

## Multimodal

- DeepSeek V4 Flash: —
- Llama 4 Maverick: 30.9 (#100)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| LMArena Vision | — | 1141 |
| GeoBench | — | 52% |
| SpatialViz-Bench | — | 31.8% |

## Multilingual

- DeepSeek V4 Flash: 53.2 (#71)
- Llama 4 Maverick: 42.4 (#190)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| LMArena Non-English | 1420 | 1269 |
| LMArena Chinese | 1467 | 1277 |
| LMArena French | 1438 | 1258 |
| LMArena German | 1418 | 1291 |
| LMArena Japanese | 1403 | 1207 |
| LMArena Korean | 1383 | 1203 |
| LMArena Russian | 1428 | 1286 |
| LMArena Spanish | 1437 | 1293 |

## Instruction Following

- DeepSeek V4 Flash: 75.1 (#76)
- Llama 4 Maverick: 71.9 (#140)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| LMArena Instruction Following | 1421 | 1267 |
| IFEval | — | 90.8% |

## Long Context

- DeepSeek V4 Flash: 43.8 (#81)
- Llama 4 Maverick: 31.4 (#274)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| LMArena Longer Query | 1435 | 1280 |
| Fiction.LiveBench | — | 46.2% |

## Writing & Preference

- DeepSeek V4 Flash: 64.0 (#59)
- Llama 4 Maverick: 39.0 (#247)

| Benchmark | DeepSeek V4 Flash | Llama 4 Maverick |
|---|---|---|
| LMArena Text | 1432 | 1287 |
| LMArena Creative Writing | 1403 | 1266 |
| EQ-Bench Creative Writing | 1559 | 860 |
| LMArena Multi-Turn | 1449 | 1289 |
| Short-Story Creative Writing | — | 62% |
| WildBench | — | 80% |

## FAQ

### Is DeepSeek V4 Flash better than Llama 4 Maverick?

DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 30.6 on the Noometry Index.

### Which is cheaper, DeepSeek V4 Flash or Llama 4 Maverick?

DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; Llama 4 Maverick lists at $0.19 and $0.65.

### Is DeepSeek V4 Flash or Llama 4 Maverick better for coding?

DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 26.3 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Flash does, with 1M tokens against 128K.

### How many benchmarks do DeepSeek V4 Flash and Llama 4 Maverick share?

32 benchmarks have published results for both models. DeepSeek V4 Flash has 41 scored results on Noometry and Llama 4 Maverick has 54.