# DeepSeek V4 Pro vs Mistral Large 3

> DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 39.1 on the Noometry Index. Mistral Large 3 costs 2.6× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/deepseek-v4-pro-vs-mistral-large-3
- Last updated: 2026-10-11
- Shared benchmarks: 22

## Summary

- They share 22 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 8 categories and Mistral Large 3 in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 15.2.
- The biggest single-benchmark swing is NYT Connections (extended): 91.3% for DeepSeek V4 Pro and 7.5% for Mistral Large 3.
- Mistral Large 3 is cheaper at $0.25 / $0.75 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 262K.

## Snapshot

| | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| Provider | DeepSeek | Mistral AI |
| Noometry Index | 54.3 | 39.1 |
| Rank | 31 | 176 |
| Context | 1M | 262K |
| Input $/M | $0.66 | $0.25 |
| Output $/M | $1.98 | $0.75 |
| Weights | Open | Open |

## Coding

- DeepSeek V4 Pro: 52.4 (#34)
- Mistral Large 3: 34.4 (#237)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena WebDev | 1582 | 1230 |
| LMArena Coding | 1470 | 1448 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| SciCode | 51% | — |
| WeirdML | 66.2% | — |
| ALE-Bench | 1,403 | — |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.8 (#58)
- Mistral Large 3: —

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| APEX-Agents | 47.3% | — |
| Vending-Bench 2 | 3,285 | — |

## Reasoning

- DeepSeek V4 Pro: 56.5 (#24)
- Mistral Large 3: 15.2 (#319)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| Kagi LLM Benchmark | 53.5% | 50.9% |
| NYT Connections (extended) | 91.3% | 7.5% |
| LMArena Hard Prompts | 1461 | 1429 |
| ARC-AGI-2 | 61.3% | — |
| ARC-AGI-1 | 90.5% | — |
| CritPt | 18% | — |
| Chess Puzzles | 47% | — |
| Thematic Generalization | — | 23% |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| Epoch Capabilities Index | 155.31 | — |
| ForecastBench | 56.1 | — |

## Math

- DeepSeek V4 Pro: 64.8 (#30)
- Mistral Large 3: 38.7 (#129)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena Math | 1455 | 1414 |
| 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: 59.5 (#31)
- Mistral Large 3: 36.0 (#177)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| Vectara Hallucination Rate | 8.6% | 14.5% |
| LMArena Expert | 1464 | 1421 |
| GPQA Diamond | 91.7% | — |
| SimpleQA Verified | 52.9% | — |

## Multimodal

- DeepSeek V4 Pro: —
- Mistral Large 3: 38.2 (#66)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena Vision | — | 1221 |

## Multilingual

- DeepSeek V4 Pro: 54.4 (#45)
- Mistral Large 3: 52.5 (#84)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena Non-English | 1439 | 1413 |
| LMArena Chinese | 1486 | 1447 |
| LMArena French | 1472 | 1455 |
| LMArena German | 1458 | 1437 |
| LMArena Japanese | 1445 | 1394 |
| LMArena Korean | 1447 | 1384 |
| LMArena Russian | 1453 | 1411 |
| LMArena Spanish | 1458 | 1440 |

## Instruction Following

- DeepSeek V4 Pro: 76.1 (#47)
- Mistral Large 3: 74.0 (#108)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1403 |

## Long Context

- DeepSeek V4 Pro: 45.0 (#51)
- Mistral Large 3: 43.1 (#105)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena Longer Query | 1458 | 1413 |
| CL-bench Life | 13.5% | — |

## Writing & Preference

- DeepSeek V4 Pro: 65.5 (#46)
- Mistral Large 3: 60.0 (#101)

| Benchmark | DeepSeek V4 Pro | Mistral Large 3 |
|---|---|---|
| LMArena Text | 1451 | 1428 |
| LMArena Creative Writing | 1446 | 1386 |
| EQ-Bench Creative Writing | 1553 | 1412 |
| LMArena Multi-Turn | 1467 | 1429 |
| EQ-Bench 4 | 1166 | — |

## FAQ

### Is DeepSeek V4 Pro better than Mistral Large 3?

DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 39.1 on the Noometry Index. Mistral Large 3 costs 2.6× 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 V4 Pro or Mistral Large 3?

Mistral Large 3 is cheaper. It lists at $0.25 per million input tokens and $0.75 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.

### Is DeepSeek V4 Pro or Mistral Large 3 better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 34.4 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Pro does, with 1M tokens against 262K.

### How many benchmarks do DeepSeek V4 Pro and Mistral Large 3 share?

22 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Mistral Large 3 has 24.