Model comparison
GPT-5.2 vs MiniMax-M3
GPT-5.2 is the stronger model overall, scoring 54.1 to 43.8 on the Noometry Index. MiniMax-M3 costs 9.2× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Last verified . 33 shared benchmarks.
MiniMaxSummary
- They share 33 benchmarks with published results for both. GPT-5.2 scores higher in 8 categories and MiniMax-M3 in 2 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 30.1.
- The biggest single-benchmark swing is Chess Puzzles: 49% for GPT-5.2 and 14% for MiniMax-M3.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- MiniMax-M3 accepts more context: 1M tokens versus 400K.
- MiniMax-M3 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.2 | MiniMax-M3 | |
|---|---|---|
| Provider | OpenAI | MiniMax |
| Noometry Index | 54.1 | 43.8 |
| Released | 2025-12-11 | 2026-06-01 |
| Weights | Proprietary | Open |
| Context window | 400K | 1M |
| Max output | 128K | 512K |
| Input $ / M tokens | $1.75 | $0.30 |
| Output $ / M tokens | $14 | $1.20 |
| Results tracked | 67 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GPT-5.2: 51.6 (#37), MiniMax-M3: 41.8 (#118)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| LMArena WebDev | 1416 | 1482 |
| LMArena Coding | 1447 | 1469 |
| ALE-Bench | 1,294 | 640.02 |
| SWE-bench Verified | 73.8% | — |
| FrontierCode | — | 14.7% |
| SWE-bench Verified (bash only) | 72.8% | — |
| SWE-bench Multilingual | 66.7% | — |
| SciCode | — | 47.1% |
| GSO | 27.4% | — |
| WeirdML | 72.2% | — |
| AlgoTune | 2.05 | — |
Agentic & Tool Use GPT-5.2 leads
GPT-5.2: 40.2 (#24), MiniMax-M3: 22.6 (#130)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| Vending-Bench 2 | 3,591 | 2,158 |
| Terminal-Bench | 64.9% | — |
| APEX-Agents | — | 37.7% |
| Berkeley Function Calling Leaderboard | 55.9% | — |
| OSWorld 2.0 | — | 4.6% |
| GDPval | 49.7% | — |
| Remote Labor Index | 2.5% | — |
| τ²-bench Airline | 83% | — |
| τ²-bench Banking | 32.2% | — |
| τ²-bench Retail | 81.6% | — |
| τ²-bench Telecom | 89.7% | — |
| DeepResearch Bench | 41.1% | — |
| GBAEval | — | 0.9% |
| LMArena Search | 1207 | — |
| METR Time Horizons | 75.3% | — |
Reasoning GPT-5.2 leads
GPT-5.2: 50.2 (#35), MiniMax-M3: 30.1 (#87)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| SimpleBench | 45.8% | 45.8% |
| NYT Connections (extended) | 83.6% | 65.1% |
| Chess Puzzles | 49% | 14% |
| LMArena Hard Prompts | 1445 | 1447 |
| Mystery Game Puzzles | 23% | 8% |
| DTBench | 90.9% | 78.9% |
| LMCA | 43.9% | 33.7% |
| Epoch Capabilities Index | 153.45 | 146.95 |
| ForecastBench | 60.1 | 61.4 |
| ARC-AGI-2 | 52.9% | — |
| Kagi LLM Benchmark | 73.3% | — |
| ARC-AGI-1 | 86.2% | — |
| CritPt | — | 3.7% |
| EnigmaEval | 10.4% | — |
| EBR-Bench | 23% | — |
| Surface Evolver Bench | — | 55% |
Math GPT-5.2 leads
GPT-5.2: 60.0 (#38), MiniMax-M3: 40.0 (#95)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 96.1% | 71.1% |
| ProofBench | 15% | 18% |
| LMArena Math | 1440 | 1429 |
| FrontierMath (Tiers 1-3) | 67.4% | — |
| FrontierMath Tier 4 | 31.7% | — |
| MathArena Final-Answer Competitions | 72% | — |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Tier 4 (v1) | 18.8% | — |
Knowledge Too close to call
GPT-5.2: 59.3 (#32), MiniMax-M3: 58.4 (#35)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 91.4% | 90.9% |
| LMArena Expert | 1445 | 1461 |
| Humanity's Last Exam | 27.8% | — |
| SimpleQA Verified | 37.1% | — |
| Vectara Hallucination Rate | 8.4% | — |
Multimodal GPT-5.2 leads
GPT-5.2: 51.3 (#7), MiniMax-M3: 40.2 (#51)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| LMArena Vision | 1268 | 1253 |
| LMArena Document | 1405 | 1435 |
| VPCT | 84% | — |
| Furniture Assembly | 38.3% | — |
Multilingual Too close to call
GPT-5.2: 53.4 (#67), MiniMax-M3: 53.0 (#75)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1425 | 1420 |
| LMArena Chinese | 1460 | 1463 |
| LMArena French | 1455 | 1447 |
| LMArena German | 1448 | 1426 |
| LMArena Japanese | 1420 | 1381 |
| LMArena Korean | 1392 | 1372 |
| LMArena Russian | 1440 | 1428 |
| LMArena Spanish | 1433 | 1432 |
Instruction Following Too close to call
GPT-5.2: 74.7 (#89), MiniMax-M3: 75.5 (#62)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1417 | 1433 |
Long Context Too close to call
GPT-5.2: 44.0 (#78), MiniMax-M3: 44.2 (#72)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1428 | 1445 |
| CL-bench | 18.2% | — |
Writing & Preference GPT-5.2 leads
GPT-5.2: 66.8 (#32), MiniMax-M3: 62.1 (#83)
| Benchmark | GPT-5.2 | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1439 | 1433 |
| LMArena Creative Writing | 1401 | 1404 |
| LMArena Multi-Turn | 1458 | 1442 |
| EQ-Bench Creative Writing | 1703 | — |
| EQ-Bench 4 | — | 1150 |
Frequently asked questions
Is GPT-5.2 better than MiniMax-M3?
GPT-5.2 is the stronger model overall, scoring 54.1 to 43.8 on the Noometry Index. MiniMax-M3 costs 9.2× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Which is cheaper, GPT-5.2 or MiniMax-M3?
MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GPT-5.2 or MiniMax-M3 better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 41.8 in the Noometry coding category.
Which has the bigger context window?
MiniMax-M3 does, with 1M tokens against 400K.
How many benchmarks do GPT-5.2 and MiniMax-M3 share?
33 benchmarks have published results for both models. GPT-5.2 has 67 scored results on Noometry and MiniMax-M3 has 41.