Model comparison
DeepSeek V4 Pro vs MiniMax-M3
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.8 on the Noometry Index. MiniMax-M3 costs 1.9× less per token, which makes it the better buy when DeepSeek V4 Pro's lead doesn't matter for your workload.
Last verified . 36 shared benchmarks.
DeepSeek
MiniMaxSummary
- They share 36 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and MiniMax-M3 in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 30.1.
- The biggest single-benchmark swing is Mystery Game Puzzles: 43% for DeepSeek V4 Pro and 8% for MiniMax-M3.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $0.66 / $1.98 for DeepSeek V4 Pro.
Side by side
| DeepSeek V4 Pro | MiniMax-M3 | |
|---|---|---|
| Provider | DeepSeek | MiniMax |
| Noometry Index | 54.3 | 43.8 |
| Released | 2026-04-24 | 2026-06-01 |
| Weights | Open | Open |
| Context window | 1M | 1M |
| Max output | 393K | 512K |
| Input $ / M tokens | $0.66 | $0.30 |
| Output $ / M tokens | $1.98 | $1.20 |
| Results tracked | 48 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), MiniMax-M3: 41.8 (#118)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| FrontierCode | 28.6% | 14.7% |
| LMArena WebDev | 1582 | 1482 |
| SciCode | 51% | 47.1% |
| LMArena Coding | 1470 | 1469 |
| ALE-Bench | 1,403 | 640.02 |
| SWE-bench Verified | 77.6% | — |
| WeirdML | 66.2% | — |
Agentic & Tool Use DeepSeek V4 Pro leads
DeepSeek V4 Pro: 32.8 (#58), MiniMax-M3: 22.6 (#130)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| APEX-Agents | 47.3% | 37.7% |
| Vending-Bench 2 | 3,285 | 2,158 |
| OSWorld 2.0 | — | 4.6% |
| GBAEval | — | 0.9% |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), MiniMax-M3: 30.1 (#87)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| NYT Connections (extended) | 91.3% | 65.1% |
| CritPt | 18% | 3.7% |
| Chess Puzzles | 47% | 14% |
| LMArena Hard Prompts | 1461 | 1447 |
| Mystery Game Puzzles | 43% | 8% |
| DTBench | 93.9% | 78.9% |
| LMCA | 45.5% | 33.7% |
| Surface Evolver Bench | 40% | 55% |
| Epoch Capabilities Index | 155.31 | 146.95 |
| ForecastBench | 56.1 | 61.4 |
| ARC-AGI-2 | 61.3% | — |
| SimpleBench | — | 45.8% |
| Kagi LLM Benchmark | 53.5% | — |
| ARC-AGI-1 | 90.5% | — |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), MiniMax-M3: 40.0 (#95)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 71.1% |
| ProofBench | 50% | 18% |
| LMArena Math | 1455 | 1429 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), MiniMax-M3: 58.4 (#35)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 91.7% | 90.9% |
| LMArena Expert | 1464 | 1461 |
| SimpleQA Verified | 52.9% | — |
| Vectara Hallucination Rate | 8.6% | — |
Multimodal Not comparable
DeepSeek V4 Pro: —, MiniMax-M3: 40.2 (#51)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| LMArena Vision | — | 1253 |
| LMArena Document | — | 1435 |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), MiniMax-M3: 53.0 (#75)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1439 | 1420 |
| LMArena Chinese | 1486 | 1463 |
| LMArena French | 1472 | 1447 |
| LMArena German | 1458 | 1426 |
| LMArena Japanese | 1445 | 1381 |
| LMArena Korean | 1447 | 1372 |
| LMArena Russian | 1453 | 1428 |
| LMArena Spanish | 1458 | 1432 |
Instruction Following Too close to call
DeepSeek V4 Pro: 76.1 (#47), MiniMax-M3: 75.5 (#62)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1448 | 1433 |
Long Context Too close to call
DeepSeek V4 Pro: 45.0 (#51), MiniMax-M3: 44.2 (#72)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1458 | 1445 |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), MiniMax-M3: 62.1 (#83)
| Benchmark | DeepSeek V4 Pro | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1451 | 1433 |
| LMArena Creative Writing | 1446 | 1404 |
| EQ-Bench 4 | 1166 | 1150 |
| LMArena Multi-Turn | 1467 | 1442 |
| EQ-Bench Creative Writing | 1553 | — |
Frequently asked questions
Is DeepSeek V4 Pro better than MiniMax-M3?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 43.8 on the Noometry Index. MiniMax-M3 costs 1.9× 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 MiniMax-M3?
MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; DeepSeek V4 Pro lists at $0.66 and $1.98.
Is DeepSeek V4 Pro or MiniMax-M3 better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 41.8 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do DeepSeek V4 Pro and MiniMax-M3 share?
36 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and MiniMax-M3 has 41.