MiniMaxMiniMax-M3
Best configuration ranks #152 of 610 on the BenchLeader Index at 56.9 ±5.0. Last measured 8 Sept 2026. Released 1 Jun 2026.
- Blended price
- $0.525/M
- $0.300 in · $1.20 out
- Output speed
- 100 tok/s
- First answer
- 22 s
- first token 1.12 s
- Context
- 1.0M
- Overall index57
- Reasoning58
- Coding53
- Agents & tools45
- Maths48
- Knowledge60
- Instruction following79
- Human preference57
- Multimodal62
- Long context68
- Composite47
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | – | – | 100 tok/s | 22 s | $0.0005 | Maths 36 · Reasoning 59 |
| defaultbest | 56.9 | #152 | 100 tok/s | 22 s | $0.0005 | Agents & tools 45 · Coding 53 · Composite 47 · Human preference 57 · Instruction following 79 · Knowledge 60 · Long context 68 · Maths 48 · Multimodal 62 · Reasoning 58 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| GPQA Diamond | 81.3%#108 | 90.9%#29 | Epoch AI Benchmarking Hub | |
| SimpleBench | – | 45.8%#52 | SimpleBench | |
| LMArena Hard Prompts | – | 1464#73 | LMArena | |
| LiveBench Reasoningnot in index | – | 74.5%#48 | LiveBench | |
| GPQA Diamond (AA)not in index | – | 92.9%#27 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | – | 39.0%#61 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | 92.7%#14 | Vals AI |
Coding
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| SciCode | – | 45.4%#85 | SciCode | |
| FrontierCode | – | 14.7%#26 | Cognition | |
| LMArena Coding | – | 1498#66 | LMArena | |
| LMArena WebDev | – | 1486#44 | LMArena | |
| LiveBench Codingnot in index | – | 68.2%#50 | LiveBench | |
| SciCode (AA)not in index | – | 47.1%#83 | Artificial Analysis | |
| LiveCodeBench | – | 82.2%#54 | Vals AI | |
| SWE-bench (Vals)not in index | – | 75.0%#42 | Vals AI |
Agents & tools
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| OSWorld-Verified 2.0 | – | 4.6%#12 | OSWorld | |
| LMArena Agent | – | -5.3#33 | LMArena | |
| LiveBench Agentic Codingnot in index | – | 40.7%#48 | LiveBench | |
| Terminal-Bench Hard | – | 42.4%#44 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | – | 88.9%#71 | Artificial Analysis | |
| Terminal-Bench 2.1 (Vals) | – | 53.6%#43 | Vals AI |
Maths
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIME | 26.7%#193 | 71.1%#120 | Epoch AI Benchmarking Hub | |
| ProofBench | – | 18.0%#38 | Vals AI | |
| LiveBench Mathematicsnot in index | – | 77.0%#51 | LiveBench |
Knowledge
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| LiveBench Data Analysisnot in index | – | 76.2%#25 | LiveBench | |
| AA-Omniscience | – | 1.4#84 | Artificial Analysis | |
| MMLU-Pro | – | 84.2%#63 | Vals AI | |
| LegalBench | – | 85.4%#18 | Vals AI | |
| CorpFin | – | 68.1%#10 | Vals AI | |
| TaxEval | – | 72.7%#58 | Vals AI |
Instruction following
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| LiveBench Languagenot in index | – | 76.8%#34 | LiveBench | |
| IFBench | – | 82.9%#3 | Artificial Analysis |
Human preference
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| LMArena Text | – | 1443#77 | LMArena | |
| EQ-Bench 4 | – | 1150#19 | EQ-Bench |
Multimodal
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| LMArena Vision | – | 1255#44 | LMArena | |
| MMMU-Pro | – | 78.5%#48 | Artificial Analysis |
Long context
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| AA-LCR | – | 83.0%#12 | Artificial Analysis |
Composite
| Benchmark | no reasoning | default | Source | Trend |
|---|---|---|---|---|
| Epoch Capabilities Indexnot in index | – | 146.5#61 | Epoch AI Benchmarking Hub | |
| LiveBench | – | 67.3%#48 | LiveBench | |
| AA Intelligence Index | – | 29.6#83 | Artificial Analysis | |
| Vals Indexnot in index | – | 42.7#33 | Vals AI |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| CoreWeave | 142 tok/s | 0.44 s | $0.230 | $0.960 | 262k | fp4 |
| SambaNova | 134 tok/s | 1.47 s | $0.600 | $2.40 | 1.0M | – |
| ModelRun [by Modular] | 102 tok/s | 1.15 s | $0.750 | $3.00 | 1.0M | fp4 |
| Venice | 83 tok/s | 1.05 s | $0.300 | $1.20 | 524k | fp8 |
| MiniMax | 77 tok/s | 0.96 s | $0.300 | $1.20 | 524k | fp8 |
| Together | 66 tok/s | 0.90 s | $0.300 | $1.20 | 524k | – |
| Parasail | 66 tok/s | 0.87 s | $0.300 | $1.20 | 1.0M | fp8 |
| NovitaAI | 63 tok/s | 1.63 s | $0.300 | $1.20 | 1M | fp8 |
| GMICloud | 59 tok/s | 1.33 s | $0.240 | $0.960 | 1.0M | fp8 |
| DeepInfra | 27 tok/s | 1.12 s | $0.280 | $1.10 | 524k | fp8 |
| StreamLake | 26 tok/s | 1.44 s | $0.300 | $1.20 | 1M | fp8 |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.060 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0005 | $0.0004 | 24.7 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0043 | $0.0022 | 27.7 s |
| Code edit | 6,000 / 1,500 | $0.0036 | $0.0025 | 36.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.023 | $0.012 | 1.0 min |
| Structured extraction | 2,000 / 200 | $0.0008 | $0.0005 | 23.7 s |
See also
Data as of 9 Sept 2026. Compare these configurations.