Model comparison
GPT-6 Astra vs MiniMax-M3
GPT-6 Astra is the stronger model overall, scoring 70.8 to 43.8 on the Noometry Index. MiniMax-M3 costs 38× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 35 shared benchmarks.
MiniMaxSummary
- They share 35 benchmarks with published results for both. GPT-6 Astra scores higher in 10 categories and MiniMax-M3 in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 30.1.
- The biggest single-benchmark swing is ProofBench: 99% for GPT-6 Astra and 18% for MiniMax-M3.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 1M.
- MiniMax-M3 has downloadable open weights; the other is API-only.
Side by side
| GPT-6 Astra | MiniMax-M3 | |
|---|---|---|
| Provider | OpenAI | MiniMax |
| Noometry Index | 70.8 | 43.8 |
| Released | 2026-09-03 | 2026-06-01 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 1M |
| Max output | 128K | 512K |
| Input $ / M tokens | $10 | $0.30 |
| Output $ / M tokens | $50 | $1.20 |
| Results tracked | 56 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GPT-6 Astra: 73.7 (#2), MiniMax-M3: 41.8 (#118)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| FrontierCode | 53.3% | 14.7% |
| LMArena WebDev | 1786 | 1482 |
| SciCode | 56.5% | 47.1% |
| LMArena Coding | 1487 | 1469 |
| ALE-Bench | 2,951 | 640.02 |
| DeepSWE | 74.1% | — |
| FrontierSWE | 65.5% | — |
| GSO | 79.4% | — |
| WeirdML | 93.6% | — |
| MirrorCode | 46.7% | — |
Agentic & Tool Use GPT-6 Astra leads
GPT-6 Astra: 52.9 (#3), MiniMax-M3: 22.6 (#130)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| APEX-Agents | 64.7% | 37.7% |
| Vending-Bench 2 | 15,515 | 2,158 |
| OSWorld 2.0 | — | 4.6% |
| Remote Labor Index | 20.8% | — |
| BALROG | 68.3% | — |
| GBAEval | — | 0.9% |
| GDP.pdf | 34.2% | — |
Reasoning GPT-6 Astra leads
GPT-6 Astra: 85.1 (#1), MiniMax-M3: 30.1 (#87)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| NYT Connections (extended) | 98.1% | 65.1% |
| CritPt | 31.7% | 3.7% |
| Chess Puzzles | 72% | 14% |
| LMArena Hard Prompts | 1462 | 1447 |
| Mystery Game Puzzles | 84% | 8% |
| DTBench | 97.3% | 78.9% |
| LMCA | 64.4% | 33.7% |
| Epoch Capabilities Index | 166.45 | 146.95 |
| ARC-AGI-2 | 95% | — |
| SimpleBench | — | 45.8% |
| ARC-AGI-1 | 98.5% | — |
| EBR-Bench | 76.2% | — |
| Surface Evolver Bench | — | 55% |
| Bench to the Future 3 | 0.14 | — |
| ForecastBench | — | 61.4 |
Math GPT-6 Astra leads
GPT-6 Astra: 93.5 (#2), MiniMax-M3: 40.0 (#95)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 100% | 71.1% |
| ProofBench | 99% | 18% |
| LMArena Math | 1465 | 1429 |
| FrontierMath (Tiers 1-3) | 93.7% | — |
| FrontierMath Tier 4 | 97.6% | — |
| FrontierMath Erdős | 2.9% | — |
Knowledge GPT-6 Astra leads
GPT-6 Astra: 75.3 (#1), MiniMax-M3: 58.4 (#35)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 95.8% | 90.9% |
| LMArena Expert | 1483 | 1461 |
| Humanity's Last Exam | 54.8% | — |
| SimpleQA Verified | 75.6% | — |
| Vectara Hallucination Rate | 8.7% | — |
Multimodal GPT-6 Astra leads
GPT-6 Astra: 55.0 (#3), MiniMax-M3: 40.2 (#51)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| LMArena Vision | 1281 | 1253 |
| LMArena Document | 1468 | 1435 |
| Blueprint-Bench 2 | 49.7% | — |
| Furniture Assembly | 80% | — |
Multilingual Too close to call
GPT-6 Astra: 53.7 (#61), MiniMax-M3: 53.0 (#75)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1430 | 1420 |
| LMArena Chinese | 1484 | 1463 |
| LMArena French | 1456 | 1447 |
| LMArena German | 1440 | 1426 |
| LMArena Japanese | 1379 | 1381 |
| LMArena Korean | 1426 | 1372 |
| LMArena Russian | 1436 | 1428 |
| LMArena Spanish | 1407 | 1432 |
Instruction Following Too close to call
GPT-6 Astra: 76.3 (#44), MiniMax-M3: 75.5 (#62)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1450 | 1433 |
Long Context Too close to call
GPT-6 Astra: 44.5 (#62), MiniMax-M3: 44.2 (#72)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1456 | 1445 |
Writing & Preference GPT-6 Astra leads
GPT-6 Astra: 75.3 (#7), MiniMax-M3: 62.1 (#83)
| Benchmark | GPT-6 Astra | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1441 | 1433 |
| LMArena Creative Writing | 1418 | 1404 |
| LMArena Multi-Turn | 1448 | 1442 |
| EQ-Bench Creative Writing | 2173 | — |
| EQ-Bench 4 | — | 1150 |
Frequently asked questions
Is GPT-6 Astra better than MiniMax-M3?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 43.8 on the Noometry Index. MiniMax-M3 costs 38× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, GPT-6 Astra or MiniMax-M3?
MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GPT-6 Astra or MiniMax-M3 better for coding?
GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 41.8 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 1M.
How many benchmarks do GPT-6 Astra and MiniMax-M3 share?
35 benchmarks have published results for both models. GPT-6 Astra has 56 scored results on Noometry and MiniMax-M3 has 41.