# DeepSeek V4 Pro vs MiniMax-M2.5

> DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 38.1 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-pro-vs-minimax-m2-5
- Last updated: 2026-10-10
- Shared benchmarks: 29

## Summary

- They share 29 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and MiniMax-M2.5 in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.4 to 17.6.
- The biggest single-benchmark swing is NYT Connections (extended): 91.3% for DeepSeek V4 Pro and 16.8% for MiniMax-M2.5.
- Both cost about the same: $0.43 input and $0.87 output per million tokens.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 205K.

## Snapshot

| | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| Provider | DeepSeek | MiniMax |
| Noometry Index | 54.2 | 38.1 |
| Rank | 28 | 181 |
| Context | 1M | 205K |
| Input $/M | $0.43 | $0.30 |
| Output $/M | $0.87 | $1.20 |
| Weights | Open | Open |

## Coding

- DeepSeek V4 Pro: 52.2 (#31)
- MiniMax-M2.5: 47.9 (#53)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| LMArena WebDev | 1582 | 1386 |
| LMArena Coding | 1470 | 1380 |
| ALE-Bench | 1,403 | 618.17 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| SWE-bench Verified (bash only) | — | 75.8% |
| SWE-bench Multilingual | — | 68.3% |
| SciCode | 51% | — |
| WeirdML | 66.2% | — |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.7 (#54)
- MiniMax-M2.5: 30.2 (#70)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| Vending-Bench 2 | 3,285 | -23.16 |
| Terminal-Bench | — | 42.7% |
| APEX-Agents | 47.3% | — |

## Reasoning

- DeepSeek V4 Pro: 56.4 (#22)
- MiniMax-M2.5: 17.6 (#281)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| ARC-AGI-2 | 61.3% | 4.9% |
| Kagi LLM Benchmark | 53.5% | 55.2% |
| NYT Connections (extended) | 91.3% | 16.8% |
| ARC-AGI-1 | 90.5% | 63.7% |
| LMArena Hard Prompts | 1461 | 1372 |
| Epoch Capabilities Index | 155.39 | 146.68 |
| CritPt | 18% | — |
| Chess Puzzles | 47% | — |
| Mystery Game Puzzles | 43% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| ForecastBench | 56.1 | — |

## Math

- DeepSeek V4 Pro: 64.6 (#24)
- MiniMax-M2.5: 25.8 (#242)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| ProofBench | 50% | 4% |
| LMArena Math | 1455 | 1378 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| OTIS Mock AIME 2024-2025 | 98.6% | — |

## Knowledge

- DeepSeek V4 Pro: 59.5 (#28)
- MiniMax-M2.5: 38.6 (#132)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| Vectara Hallucination Rate | 8.6% | 9.1% |
| LMArena Expert | 1463 | 1378 |
| GPQA Diamond | 91.7% | — |
| SimpleQA Verified | 52.9% | — |

## Multilingual

- DeepSeek V4 Pro: 54.5 (#44)
- MiniMax-M2.5: 47.3 (#147)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| LMArena Non-English | 1438 | 1338 |
| LMArena Chinese | 1486 | 1392 |
| LMArena French | 1473 | 1363 |
| LMArena German | 1458 | 1361 |
| LMArena Japanese | 1448 | 1169 |
| LMArena Korean | 1452 | 1232 |
| LMArena Russian | 1452 | 1359 |
| LMArena Spanish | 1457 | 1354 |

## Instruction Following

- DeepSeek V4 Pro: 76.3 (#44)
- MiniMax-M2.5: 71.7 (#143)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1353 |

## Long Context

- DeepSeek V4 Pro: 45.0 (#49)
- MiniMax-M2.5: 37.4 (#211)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| CL-bench Life | 13.5% | 6.3% |
| LMArena Longer Query | 1459 | 1367 |
| CL-bench | — | 11.4% |

## Writing & Preference

- DeepSeek V4 Pro: 65.6 (#45)
- MiniMax-M2.5: 54.1 (#149)

| Benchmark | DeepSeek V4 Pro | MiniMax-M2.5 |
|---|---|---|
| LMArena Text | 1451 | 1359 |
| LMArena Creative Writing | 1446 | 1331 |
| EQ-Bench Creative Writing | 1553 | 1361 |
| LMArena Multi-Turn | 1467 | 1364 |
| EQ-Bench 4 | 1166 | — |

## FAQ

### Is DeepSeek V4 Pro better than MiniMax-M2.5?

DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 38.1 on the Noometry Index.

### Which is cheaper, DeepSeek V4 Pro or MiniMax-M2.5?

MiniMax-M2.5 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; DeepSeek V4 Pro lists at $0.43 and $0.87.

### Is DeepSeek V4 Pro or MiniMax-M2.5 better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.2 versus 47.9 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Pro does, with 1M tokens against 205K.

### How many benchmarks do DeepSeek V4 Pro and MiniMax-M2.5 share?

29 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and MiniMax-M2.5 has 33.