Home/Models
Models/MiniMax (minimax.io)

MiniMax-M3

MiniMax-M3 is a multimodal foundation model from MiniMax. It supports text, image, and video inputs with text output, a 1M-token context window, and is suited for long-horizon agentic work, coding, and tool use. It is built on MiniMax Sparse Attention (MSA), which replaces full attention with KV-block selection to cut per-token compute at long context — roughly 1/20 the cost of the previous generation at 1M tokens, with substantially faster prefill and decode while retaining quality across most tasks.

Trained as a native multimodal model on interleaved data and tuned for multi-turn, production-like collaboration via an interactive user-simulator framework, the model is oriented toward sustained, multi-step tasks rather than single-turn execution.

minimax/MiniMax-M3
Organization
MiniMax (minimax.io)
Family
minimax
Providers
51
Context
1,048,576
Output limit
512,000
Knowledge
—
Release
2026-06-01
Updated
—
Weights
Open
Input
text, image/jpeg, image/png, image/gif, image/webp, video/mp4, video/x-msvideo, video/quicktime, video/x-matroska
Output
text
Capabilities
—

Providers

Compare every cataloged offer for this model, including its provider, route, token limits, price, and capabilities.

71 offers · USD per million tokens
ProviderModel IDContextOutputInput / output · 1MReasoningToolsStructuredDetails
302.AI
MiniMax-M3
1,048,576512,000— / ————
View
Abacus
MiniMaxAI/MiniMax-M3
1,000,000131,072— / ————
View
AIHubMix
minimax-m3
1,000,000512,000— / ————
View
Ambient
minimax/minimax-m3
131,072131,702— / ————
View
Boundless
minimax-m3
1,048,576512,000— / ————
View
Charm Hyper
minimax-m3
512,000512,000— / ————
View
ClinePass
cline-pass/minimax-m3
1,048,576512,000— / ————
View
CommandCode
MiniMaxAI/MiniMax-M3
1,000,000512,000— / ————
View
CoreWeave
MiniMaxAI/MiniMax-M3
262,144262,144— / ————
View
Cortecs
minimax-m3
1,048,5761,048,576— / ————
View
CrossModel
minimax/minimax-m3
1,024,000512,000— / ————
View
Deep Infra
MiniMaxAI/MiniMax-M3
524,288512,000— / ————
View
DevPass (LLM Gateway)
minimax-m3
1,048,576512,000— / ————
View
Eden AI
minimax/MiniMax-M3
524,288512,000— / ————
View
EmpirioLabs AI
minimax-m3
1,000,000524,288— / ————
View
Fireworks AI
accounts/fireworks/models/minimax-m3
512,000512,000— / ————
View
Fireworks AI
accounts/fireworks/routers/minimax-latest
512,000512,000— / ————
View
GMI Cloud
MiniMaxAI/MiniMax-M3
1,048,576512,000— / ————
View
Hugging Face
MiniMaxAI/MiniMax-M3
524,288512,000— / ————
View
Inco
minimax-m3:fast
1,048,576512,000— / ————
View
Inco
minimax-m3
1,048,576512,000— / ————
View
IteraCompute
minimax/minimax-m3
1,048,576524,288— / ————
View
Jalapeno Cloud
MiniMax-M3
524,288512,000— / ————
View
Kenari
minimax-m3
1,048,576512,000— / ————
View
Kilo Gateway
minimax/minimax-m3
524,288512,000— / ————
View
Lilac
minimaxai/minimax-m3
1,048,5761,048,576— / ————
View
LLM Gateway
together-ai/minimax-m3
524,288131,072— / ————
View
LLM Gateway
tencent/minimax-m3
1,000,000131,072— / ————
View
LLM Gateway
minimax/minimax-m3
512,000131,072— / ————
View
LLM Gateway
consensusprotocol/minimax-m3
1,048,576512,000— / ————
View
Merge Gateway
minimax/minimax-m3
1,000,000128,000— / ————
View
NanoGPT
minimax/minimax-m3
512,00080,000— / ————
View
NanoGPT
minimax/minimax-m3:thinking
512,00080,000— / ————
View
Nebius Token Factory
MiniMaxAI/MiniMax-M3
1,048,5761,048,576— / ————
View
Novita AI
minimax/minimax-m3
1,000,000131,072— / ————
View
Ofox
minimax/minimax-m3
1,048,576512,000— / ————
View
Ollama Cloud
minimax-m3
512,000131,072— / ————
View
OpenCode
minimax-m3-free
200,00032,000— / ————
View
OpenCode
minimax-m3
512,000128,000— / ————
View
OpenCode Go
minimax-m3
1,000,000131,072— / ————
View
AtlasCloudvia OpenRouter
minimax/minimax-m3
fp8		524,300524,288$0.30 / $1.20YesYesNo
View
CoreWeavevia OpenRouter
minimax/minimax-m3
fp4		262,144235,929$0.23 / $0.96YesYesYes
View
DeepInfravia OpenRouter
minimax/minimax-m3
fp8		524,288131,072$0.28 / $1.10YesYesNo
View
GMICloudvia OpenRouter
minimax/minimax-m3
fp8		1,048,576524,288$0.24 / $0.96YesYesNo
View
Maravia OpenRouter
minimax/minimax-m3
unknown		1,048,576943,718$0.60 / $2.40YesYesNo
View
Minimaxvia OpenRouter
minimax/minimax-m3
fp8		524,288512,000$0.30 / $1.20YesYesNo
View
ModelRunvia OpenRouter
minimax/minimax-m3
fp4		1,048,576943,718$0.75 / $3.00YesYesYes
View
Novitavia OpenRouter
minimax/minimax-m3
fp8		1,000,000131,072$0.30 / $1.20YesYesNo
View
Parasailvia OpenRouter
minimax/minimax-m3
fp8		1,048,576524,288$0.30 / $1.20YesYesNo
View
SambaNovavia OpenRouter
minimax/minimax-m3
unknown		1,048,576943,718$0.60 / $2.40YesYesNo
View
StreamLakevia OpenRouter
minimax/minimax-m3
fp8		1,000,000512,000$0.30 / $1.20YesNoNo
View
Togethervia OpenRouter
minimax/minimax-m3
unknown		524,288471,859$0.30 / $1.20YesYesYes
View
Venicevia OpenRouter
minimax/minimax-m3
fp8		524,28865,536$0.30 / $1.20YesYesNo
View
Opper
minimax-m3
1,000,000131,072— / ————
View
OrcaRouter
minimax/minimax-m3
1,048,576512,000— / ————
View
Requesty
minimax-m3
1,000,000128,000— / ————
View
Requesty
minimax-m3@eu
1,048,5761,048,576— / ————
View
SambaNova
MiniMax-M3
1,048,5761,048,576— / ————
View
SCNet Token Plan
MiniMax-M3
1,048,576512,000— / ————
View
SiliconFlow
MiniMaxAI/MiniMax-M3
1,048,576131,000— / ————
View
Standard Compute
minimax/minimax-m3
1,048,57650,000— / ————
View
Synthetic
hf:MiniMaxAI/MiniMax-M3
524,28865,536— / ————
View
Tempr Gateway
minimax/MiniMax-M3
1,048,576512,000— / ————
View
TensorX
minimax/minimax-m3
1,048,576131,072— / ————
View
Together AI
MiniMaxAI/MiniMax-M3
524,288524,288— / ————
View
Vancine
MiniMax-M3
1,048,576512,000— / ————
View
Vercel AI Gateway
minimax/minimax-m3
512,000512,000— / ————
View
Volcengine Ark Coding Plan
minimax-m3
1,048,576512,000— / ————
View
Vultr
minimax-m3
524,288512,000— / ————
View
Wafer
MiniMax-M3
1,048,576512,000— / ————
View
ZenMux
minimax/minimax-m3
1,000,000512,000— / ————
View

Prices and limits apply to each offer. A dash means the value is not provided. OpenRouter offers show the hosting provider and routing channel separately.

Pricing

Compare token rates and cache charges across available routes and hosting configurations.

USD per million tokens
Input from$0.23Per 1M input tokens
Output from$0.96Per 1M output tokens
Available offers71Across 51 providers

Lowest listed input and output prices across the offers above; they may belong to different providers. These are base rates, before conditional pricing, discounts or additional fees.

Endpoint price comparisonLatest observed snapshot · 10 lowest-priced configurations
InputOutput
CoreWeavecoreweave/fp4
$0.23$0.96
GMICloudgmicloud/fp8
$0.24$0.96
DeepInfradeepinfra/fp8
$0.28$1.10
AtlasCloudatlas-cloud/fp8
$0.30$1.20
Minimaxminimax/fp8
$0.30$1.20
Novitanovita/fp8
$0.30$1.20
Parasailparasail/fp8
$0.30$1.20
StreamLakestreamlake/fp8
$0.30$1.20
Togethertogether
$0.30$1.20
Venicevenice/fp8
$0.30$1.20
minimax/minimax-m3Observed
Input
$0.30
Output
$1.20
Cache read
$0.06
Hosting prices through OpenRouter All endpoints and cache rates
EndpointInput / 1MOutput / 1MCache read / 1MCache write / 1M
AtlasCloudatlas-cloud/fp8$0.30$1.20$0.06—
CoreWeavecoreweave/fp4$0.23$0.96$0.05—
DeepInfradeepinfra/fp8$0.28$1.10$0.056—
GMICloudgmicloud/fp8$0.24$0.96$0.048—
Maramara$0.60$2.40——
Minimaxminimax/fp8$0.30$1.20$0.06—
ModelRunmodelrun/fp4$0.75$3.00$0.15—
Novitanovita/fp8$0.30$1.20$0.06—
Parasailparasail/fp8$0.30$1.20$0.06—
SambaNovasambanova$0.60$2.40$0.06—
StreamLakestreamlake/fp8$0.30$1.20$0.06—
Togethertogether$0.30$1.20$0.06—
Venicevenice/fp8$0.30$1.20$0.06—

Each row is a hosting configuration. Open an offer in Providers for its full pricing conditions and additional fees. Endpoint observations may differ from the routing catalog's base price.

Performance

Review reported latency and output-throughput percentiles from the latest observed rolling window.

Latest 30-minute observation

Time to first token · milliseconds

Latency distributionRolling 30-minute snapshot · 6 endpoints
P50P75P90P99
AtlasCloudatlas-cloud/fp8
1,005–10,455.92 ms
CoreWeavecoreweave/fp4
775–7,821.98 ms
DeepInfradeepinfra/fp8
1,142.5–17,485.72 ms
GMICloudgmicloud/fp8
2,233–7,435.74 ms
Maramara
1,051–36,536.5 ms
Minimaxminimax/fp8
1,120–6,332.93 ms
All latency percentiles 13 endpoints
EndpointP50P75P90P99
AtlasCloudatlas-cloud/fp81,005 ms1,393 ms2,413.6 ms10,455.92 ms
CoreWeavecoreweave/fp4775 ms1,271.5 ms2,566.3 ms7,821.98 ms
DeepInfradeepinfra/fp81,142.5 ms2,562.5 ms4,325.6 ms17,485.72 ms
GMICloudgmicloud/fp82,233 ms3,125 ms3,877.2 ms7,435.74 ms
Maramara1,051 ms1,244 ms2,544 ms36,536.5 ms
Minimaxminimax/fp81,120 ms1,637 ms2,673.9 ms6,332.93 ms
ModelRunmodelrun/fp41,777 ms2,016 ms2,814 ms7,435.2 ms
Novitanovita/fp82,009 ms4,313 ms6,089.6 ms15,392.8 ms
Parasailparasail/fp8790 ms1,560.75 ms3,406 ms18,551.5 ms
SambaNovasambanova1,684 ms2,534.5 ms5,239.6 ms23,855.71 ms
StreamLakestreamlake/fp81,384 ms1,973 ms3,583.4 ms9,762.36 ms
Togethertogether687.5 ms1,059 ms1,880.8 ms5,147.37 ms
Venicevenice/fp8926 ms1,273 ms1,694 ms11,946.95 ms

Output throughput · tokens per second

Throughput distributionRolling 30-minute snapshot · 6 endpoints
P50P75P90P99
AtlasCloudatlas-cloud/fp8
87–200.25 t/s
CoreWeavecoreweave/fp4
79–200.54 t/s
DeepInfradeepinfra/fp8
15–37 t/s
GMICloudgmicloud/fp8
66–177.51 t/s
Maramara
81–231.02 t/s
Minimaxminimax/fp8
83–199 t/s
All throughput percentiles 13 endpoints
EndpointP50P75P90P99
AtlasCloudatlas-cloud/fp887 t/s112 t/s130.5 t/s200.25 t/s
CoreWeavecoreweave/fp479 t/s103 t/s130 t/s200.54 t/s
DeepInfradeepinfra/fp815 t/s19 t/s23.8 t/s37 t/s
GMICloudgmicloud/fp866 t/s91.75 t/s118 t/s177.51 t/s
Maramara81 t/s138.5 t/s185 t/s231.02 t/s
Minimaxminimax/fp883 t/s109 t/s139 t/s199 t/s
ModelRunmodelrun/fp4111 t/s153 t/s185 t/s237.39 t/s
Novitanovita/fp876 t/s97 t/s126.6 t/s183.32 t/s
Parasailparasail/fp8129.5 t/s180.5 t/s222.6 t/s271 t/s
SambaNovasambanova37 t/s92.75 t/s104.9 t/s121 t/s
StreamLakestreamlake/fp876 t/s94 t/s114.4 t/s150.96 t/s
Togethertogether74 t/s100 t/s127 t/s168.24 t/s
Venicevenice/fp874 t/s143 t/s172.6 t/s232.58 t/s

Percentiles describe the observed request distribution. These are snapshots of a rolling window, not a historical time series.

Uptime

Compare successful request rates across the latest 5-minute, 30-minute, and 24-hour observations.

Successful requests by window
Availability by endpointCurrent overlapping windows · 13 endpoints
Endpoint5 minutes30 minutes24 hours
AtlasCloudatlas-cloud/fp8
CoreWeavecoreweave/fp4
DeepInfradeepinfra/fp8
GMICloudgmicloud/fp8
Maramara
Minimaxminimax/fp8
ModelRunmodelrun/fp4
Novitanovita/fp8
Parasailparasail/fp8
SambaNovasambanova
StreamLakestreamlake/fp8
Togethertogether
Venicevenice/fp8

Reported by OpenRouter at the time of collection; rate-limited requests are excluded. A dash means no measurement was reported. This is not a live availability check.

Benchmarks

Review attributed evaluation results on each source's original scale.

External evaluations
minimax/minimax-m3Observed
Benchmark profileArtificial Analysis · shared source scale from 0 to 60
Coding58.6
Agentic29.5
Intelligence29.2
Design ArenaCategoryEloWin rateRank
agentsagenticgamedev1,15844.6%19
agentsandroidnative1,17144.3%23
agentsfullstack1,19647.3%19
agentshtmlslides1,18046.6%12
agentsmobileapps1,16746%23
agentspython-pptxslides1,20446.8%14
agentswebapps1,19247.1%21
models3d1,21450.2%41
modelsasciiart1,16946.8%34
modelscodecategories1,25150.9%35
modelsdataviz1,23951.3%36
modelsgamedev1,22046%43
modelssvg1,16245.2%38
modelsuicomponent1,25050.9%37
modelswebsite1,26151.9%34

Sources: Artificial Analysis and Design Arena, via OpenRouter. Scores retain their original scales and are not OpenAI Suite ratings.

Detailed evaluations

Evaluation scoresAccuracy or item F1 · higher is better
tau bench verified airlineMiniMax: MiniMax M3
75.152%
gpqa diamondMiniMax: MiniMax M3
90.462%
EvaluationConfigurationSourceResultsDetails
models · websiteMiniMax M3minimax/minimax-m3-20260531Design Arena1,256 Elo
Details
models · codecategoriesMiniMax M3minimax/minimax-m3-20260531Design Arena1,246 Elo
Details
models · gamedevMiniMax M3minimax/minimax-m3-20260531Design Arena1,212 Elo
Details
models · datavizMiniMax M3minimax/minimax-m3-20260531Design Arena1,236 Elo
Details
tau bench verified airlineMiniMax: MiniMax M3minimax/minimax-m3-20260531OpenRouter75.152%
Details
models · 3dMiniMax M3minimax/minimax-m3-20260531Design Arena1,210 Elo
Details
Composite indicesMiniMax-M3minimax/minimax-m3-20260531Artificial AnalysisIntelligence 29.2 · Coding 58.6 · Agentic 29.5
Details
models · svgMiniMax M3minimax/minimax-m3-20260531Design Arena1,160 Elo
Details
models · uicomponentMiniMax M3minimax/minimax-m3-20260531Design Arena1,242 Elo
Details
gpqa diamondMiniMax: MiniMax M3minimax/minimax-m3-20260531OpenRouter90.462%
Details
models · asciiartMiniMax M3minimax/minimax-m3-20260531Design Arena1,161 Elo
Details

Sources: Artificial Analysis, Design Arena and OpenRouter. Evaluation configurations and scales differ; results are not interchangeable. Updated Oct 24, 2026.

Apps & session costs

Compare observed median session costs for applications using this model, grouped by conversation length.

30-day sample · through Oct 18, 2026
ApplicationModel routeTurnsMedian cost / sessionDetails
Claude Codeminimax/minimax-m3-2026053110–49 turns
$0.152378
Details
Kilo Codeminimax/minimax-m3-202605312–9 turns
$0.017276
Details
Kilo Codeminimax/minimax-m3-2026053110–49 turns
$0.087187
Details
Hermes Agentminimax/minimax-m3-202605312–9 turns
$0.010754
Details
Claude Codeminimax/minimax-m3-202605312–9 turns
$0.018285
Details
Hermes Agentminimax/minimax-m3-2026053110–49 turns
$0.094797
Details
Hermes Agentminimax/minimax-m3-2026053150+ turns
$0.536929
Details
Hermes Agentminimax/minimax-m3-202605311 turn
$0.002954
Details

Source: OpenRouter, as of Oct 20, 2026. Published under CC BY 4.0. These are observed costs per session, not token prices or a forecast for your workload. Apps are compared separately.

Activity

Daily prompt and completion tokens reported for this model's routes in OpenRouter's top 50. Missing days have no published value.

Last 90 complete UTC days
Daily token usage19,023,404,361,981 reported tokens · 89 days with observations
Jul 26, 2026Oct 23, 2026
Daily observations · 89 days
Date (UTC)Reported tokensReported routes
Oct 23, 2026209,609,491,3451
Oct 22, 2026186,907,057,4191
Oct 21, 2026181,136,327,0111
Oct 20, 2026194,072,869,6391
Oct 19, 2026206,483,428,1601
Oct 18, 2026174,060,867,6601
Oct 17, 2026179,124,472,6121
Oct 16, 2026210,130,855,5331
Oct 15, 2026208,245,074,8991
Oct 14, 2026205,834,854,0361
Oct 13, 2026214,019,237,8141
Oct 12, 2026202,523,712,9361
Oct 11, 2026169,862,382,9761
Oct 10, 2026179,003,985,0961
Oct 9, 2026208,885,357,6051
Oct 8, 2026201,350,291,2801
Oct 7, 2026193,845,922,2291
Oct 6, 2026176,424,919,9921
Oct 5, 2026161,872,438,7111
Oct 4, 2026138,542,150,9961
Oct 3, 2026153,677,391,7201
Oct 2, 2026157,866,702,8281
Oct 1, 2026163,189,980,3841
Sep 30, 2026189,196,211,7651
Sep 29, 2026195,423,286,0821
Sep 28, 2026189,379,268,3381
Sep 27, 2026175,384,412,4051
Sep 26, 2026161,369,136,4181
Sep 25, 2026174,704,255,7201
Sep 24, 2026210,187,157,9821
Sep 23, 2026224,857,441,6561
Sep 22, 2026203,298,904,4521
Sep 21, 2026219,164,499,7791
Sep 20, 2026179,790,059,9781
Sep 19, 2026182,701,097,6261
Sep 18, 2026218,835,420,6731
Sep 17, 2026238,209,982,5571
Sep 16, 2026233,110,589,1871
Sep 15, 2026222,635,815,4841
Sep 14, 2026209,770,398,2731
Sep 13, 2026174,851,444,4821
Sep 12, 2026180,952,746,3321
Sep 11, 2026204,939,970,0321
Sep 10, 2026233,746,793,6071
Sep 9, 2026219,600,776,6651
Sep 8, 2026225,769,208,0021
Sep 7, 2026220,962,878,4701
Sep 6, 2026178,955,132,3601
Sep 5, 2026186,672,879,5931
Sep 4, 2026211,383,224,4571
Sep 3, 2026210,540,711,3351
Sep 2, 2026220,875,513,5891
Sep 1, 2026222,563,108,0921
Aug 31, 2026215,593,028,9561
Aug 30, 2026197,812,685,4351
Aug 29, 2026180,907,822,0861
Aug 28, 2026200,251,782,8851
Aug 27, 2026238,219,507,7801
Aug 26, 2026247,266,600,0291
Aug 25, 2026239,651,565,2011
Aug 24, 2026212,511,149,3731
Aug 23, 2026175,841,459,4941
Aug 22, 2026180,786,291,1151
Aug 21, 2026199,350,164,9411
Aug 20, 2026243,364,539,2481
Aug 19, 2026314,903,730,8171
Aug 18, 2026334,704,875,3771
Aug 17, 2026252,520,942,5371
Aug 16, 2026205,819,592,0941
Aug 15, 2026206,850,067,6201
Aug 14, 2026223,992,834,3601
Aug 13, 2026212,846,920,2691
Aug 12, 2026234,129,475,6041
Aug 11, 2026211,821,249,7891
Aug 10, 2026236,731,781,5011
Aug 9, 2026207,614,344,1801
Aug 8, 2026224,604,032,8741
Aug 7, 2026234,298,378,4081
Aug 6, 2026258,564,506,0111
Aug 5, 2026270,192,133,2911
Aug 4, 2026254,794,567,7141
Aug 3, 2026253,462,188,5921
Aug 2, 2026215,099,253,8381
Aug 1, 2026234,733,554,9851
Jul 31, 2026276,309,326,2821
Jul 30, 2026304,003,653,2721
Jul 29, 2026304,597,064,7481
Jul 28, 2026320,356,235,3221
Jul 27, 2026302,394,957,6811

Source: OpenRouter (openrouter.ai/rankings), as of Oct 24, 2026. CC BY 4.0. Totals include only individually published routes. Tokenizers differ by provider; the aggregated “other” category is never assigned to a model.

Technical details & API

Explore specifications, supported parameters, reasoning controls, and API identifiers for each OpenRouter route.

Current route specifications
minimax/minimax-m3Observed
Context
1,048,576
Maximum output
512,000
Tokenizer
Other
Input
text, image, video
Output
text
Moderation
No
Added to OpenRouter
2026-05-31
Base URLhttps://openrouter.ai/api/v1
Model IDminimax/minimax-m3
Version IDminimax/minimax-m3-20260531
Weights IDMiniMaxAI/Minimax-M3

Reasoning

Required
No
Enabled by default
Not reported
Default effort
—
Supported efforts
—

Supported parameters

  • frequency_penalty
  • include_reasoning
  • logit_bias
  • logprobs
  • max_tokens
  • min_p
  • presence_penalty
  • reasoning
  • repetition_penalty
  • response_format
  • seed
  • stop
  • structured_outputs
  • temperature
  • tool_choice
  • tools
  • top_k
  • top_logprobs
  • top_p

Default parameters

top p
0.95
temperature
1

Frequently asked questions

Answers based on the model metadata and serving offers currently in the catalog.

How much does MiniMax-M3 cost?

Listed rates start at $0.23 per million input tokens and $0.96 per million output tokens. Prices vary by provider and configuration; see Pricing for details.

Which providers offer MiniMax-M3?

There are 71 cataloged offers across 51 providers. Compare model IDs, prices and limits in Providers.

What is the context window?

The cataloged model context is 1,048,576 tokens. Each serving endpoint may apply a different limit.

Does it support tools and structured output?

Tool calling: Not reported. Structured output: Not reported. Support can vary by endpoint.

Resources

Hugging Face ↗
Catalog history · 3 recorded revisions

Metadata observations since this model was first cataloged. These are not model release versions.

  • · 832ccbfe53bc
  • · af68047d4295
  • · 66d988254ce3