Model comparison
DeepSeek-V3.2-Exp vs MiniMax-M3
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 43.5 on the Noometry Index.
Last verified . 29 shared benchmarks.
DeepSeek
MiniMaxSummary
- They share 29 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 5 categories and MiniMax-M3 in 4 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where DeepSeek-V3.2-Exp leads 37.9 to 22.5.
- The biggest single-benchmark swing is NYT Connections (extended): 36.7% for DeepSeek-V3.2-Exp and 65.1% for MiniMax-M3.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.30 / $1.20 for MiniMax-M3.
- MiniMax-M3 accepts more context: 1M tokens versus 164K.
Side by side
| DeepSeek-V3.2-Exp | MiniMax-M3 | |
|---|---|---|
| Provider | DeepSeek | MiniMax |
| Noometry Index | 44.9 | 43.5 |
| Released | 2025-09-29 | 2026-06-01 |
| Weights | Open | Open |
| Context window | 164K | 1M |
| Max output | 66K | 512K |
| Input $ / M tokens | $0.26 | $0.30 |
| Output $ / M tokens | $0.38 | $1.20 |
| Results tracked | 48 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 46.4 (#61), MiniMax-M3: 41.5 (#111)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| LMArena WebDev | 1362 | 1482 |
| SciCode | 38.9% | 47.1% |
| LMArena Coding | 1454 | 1470 |
| FrontierCode | — | 14.7% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| WeirdML | 39.5% | — |
| ALE-Bench | — | 640.02 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 37.9 (#31), MiniMax-M3: 22.5 (#121)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| Vending-Bench 2 | 1,034 | 2,158 |
| Terminal-Bench | 39.6% | — |
| APEX-Agents | — | 37.7% |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| OSWorld 2.0 | — | 4.6% |
| TheAgentCompany | 42.9% | — |
| GBAEval | — | 0.9% |
Reasoning MiniMax-M3 leads
DeepSeek-V3.2-Exp: 22.0 (#185), MiniMax-M3: 29.3 (#83)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| NYT Connections (extended) | 36.7% | 65.1% |
| CritPt | 2.9% | 3.7% |
| Chess Puzzles | 14% | 14% |
| LMArena Hard Prompts | 1433 | 1447 |
| DTBench | 87.7% | 78.9% |
| LMCA | 29.1% | 33.7% |
| Epoch Capabilities Index | 146.28 | 147 |
| ARC-AGI-2 | 4% | — |
| SimpleBench | — | 45.8% |
| Kagi LLM Benchmark | 52.2% | — |
| ARC-AGI-1 | 57% | — |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 8% |
| Surface Evolver Bench | — | 55% |
| ForecastBench | — | 61.4 |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 40.8 (#75), MiniMax-M3: 39.1 (#83)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 71.1% |
| ProofBench | 8% | 18% |
| LMArena Math | 1435 | 1432 |
| MathArena Final-Answer Competitions | 57.7% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge MiniMax-M3 leads
DeepSeek-V3.2-Exp: 51.6 (#61), MiniMax-M3: 58.4 (#32)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 83.4% | 90.9% |
| LMArena Expert | 1436 | 1462 |
| Vectara Hallucination Rate | 5.3% | — |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, MiniMax-M3: 39.5 (#45)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| LMArena Vision | — | 1254 |
| LMArena Document | — | 1435 |
Multilingual Too close to call
DeepSeek-V3.2-Exp: 52.4 (#85), MiniMax-M3: 53.2 (#70)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1409 | 1420 |
| LMArena Chinese | 1461 | 1465 |
| LMArena French | 1435 | 1445 |
| LMArena German | 1440 | 1424 |
| LMArena Japanese | 1377 | 1381 |
| LMArena Korean | 1372 | 1372 |
| LMArena Russian | 1423 | 1430 |
| LMArena Spanish | 1441 | 1431 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.7 (#89), MiniMax-M3: 75.7 (#56)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1433 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), MiniMax-M3: 44.2 (#68)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1428 | 1445 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference Too close to call
DeepSeek-V3.2-Exp: 62.5 (#73), MiniMax-M3: 62.3 (#79)
| Benchmark | DeepSeek-V3.2-Exp | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1425 | 1433 |
| LMArena Creative Writing | 1403 | 1405 |
| LMArena Multi-Turn | 1427 | 1441 |
| EQ-Bench Creative Writing | 1515 | — |
| EQ-Bench 4 | — | 1150 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than MiniMax-M3?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 43.5 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or MiniMax-M3?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; MiniMax-M3 lists at $0.30 and $1.20.
Is DeepSeek-V3.2-Exp or MiniMax-M3 better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 41.5 in the Noometry coding category.
Which has the bigger context window?
MiniMax-M3 does, with 1M tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and MiniMax-M3 share?
29 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and MiniMax-M3 has 41.