BenchLeader
Menu
LeaderboardsModelsProvidersBenchmarksCompareReleasesBlogMethodology
MiniMaxOpen weights: MiniMaxAI/MiniMax-M2.7Reasoning modelAuto-detected

MiniMax-M2.7

Best configuration ranks #156 of 610 on the BenchLeader Index at 56.5 ±5.3. Last measured 8 Sept 2026. Released 18 Mar 2026.

Blended price
$0.525/M
$0.300 in · $1.20 out
Output speed
62 tok/s
First answer
41 s
first token 1.37 s
Context
205k
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index57
  2. Reasoning61
  3. Coding50
  4. Agents & tools46
  5. Maths48
  6. Knowledge55
  7. Instruction following73
  8. Human preference60
  9. Long context66
  10. Composite58

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

BenchmarkdefaultSourceTrend
LMArena Hard Prompts1440#107LMArena
GPQA Diamond (AA)not in index87.4%#98Artificial Analysis
Humanity's Last Exam (AA)not in index29.6%#106Artificial Analysis
GPQA Diamond (Vals)not in index86.6%#39Vals AI

Coding

BenchmarkdefaultSourceTrend
SciCode47.0%#77SciCode
WeirdML37.0%#117WeirdML
LMArena Coding1480#91LMArena
LMArena WebDev1398#71LMArena
SciCode (AA)not in index50.1%#68Artificial Analysis
LiveCodeBench79.9%#69Vals AI
IOI4.9%#43Vals AI
SWE-bench (Vals)not in index73.8%#49Vals AI

Agents & tools

BenchmarkdefaultSourceTrend
Terminal-Bench45.1%#31Terminal-Bench
LMArena Agent-13#39LMArena
Terminal-Bench Hard39.4%#56Artificial Analysis
τ²-Bench Telecom (AA)not in index84.8%#95Artificial Analysis
Terminal-Bench 2.1 (Vals)48.7%#51Vals AI

Maths

BenchmarkdefaultSourceTrend
ProofBench3.0%#58Vals AI
AIME (Vals)91.0%#26Vals AI

Knowledge

BenchmarkdefaultSourceTrend
AA-Omniscience0.8#90Artificial Analysis
MMLU-Pro80.4%#85Vals AI
LegalBench84.0%#40Vals AI
CorpFin61.2%#56Vals AI
TaxEval66.6%#107Vals AI

Instruction following

BenchmarkdefaultSourceTrend
IFBench75.7%#27Artificial Analysis

Human preference

BenchmarkdefaultSourceTrend
LMArena Text1415#111LMArena

Long context

BenchmarkdefaultSourceTrend
AA-LCR78.3%#77Artificial Analysis

Composite

BenchmarkdefaultSourceTrend
Epoch Capabilities Indexnot in index145.8#66Epoch AI Benchmarking Hub
AA Intelligence Index23.2#134Artificial Analysis
Vals Indexnot in index24.6#45Vals AI

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.

ProviderSpeedFirst tokenInput $/MOutput $/MContextQuantisation
Groq237 tok/s0.81 s$0.600$1.80197k–
DeepInfra80 tok/s0.61 s$0.250$1.00197kfp8
MARA75 tok/s8.04 s$0.240$0.960197k–
MiniMax Highspeed47 tok/s1.21 s$0.600$2.40205kfp8
MiniMax46 tok/s2.29 s$0.300$1.20205kfp8
NovitaAI42 tok/s1.31 s$0.270$1.08205kfp8
SambaNova41 tok/s1.85 s$0.600$2.40197k–
AtlasCloud32 tok/s1.37 s$0.300$1.20197kfp8
GMICloud30 tok/s2.89 s$0.210$0.840197kfp8

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

$0.00$0.330$0.660$0.990$1.32Jul 26Aug 26Sept 26
input outputnow $0.210 in · $0.840 out

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.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0005$0.000446.1 s
Summarise a 30-page report12,000 / 600$0.0043$0.002250.9 s
Code edit6,000 / 1,500$0.0036$0.00251.1 min
Agentic coding session60,000 / 4,000$0.023$0.0121.8 min
Structured extraction2,000 / 200$0.0008$0.000544.5 s

See also

Best under $2/MBest open weightsBest for codingBest for agentsAll MiniMax models

Data as of 9 Sept 2026. Compare with another model.