MiMo-V2.6-Pro - Intelligence, Performance & Price Analysis | Artificial AnalysisArtificial AnalysisKArtificial AnalysisModelsCoding AgentsImage, Speech, VideoInferenceLeaderboardsAboutAI TrendsArenasKXiaomi•Open weights model•Released September 2026MiMo-V2.6-Pro Intelligence, Performance & Price AnalysisCompareTry it out API Provider Benchmarks Model summaryIntelligenceUpdated#1 / 11446Artificial Analysis Intelligence Index4 out of 4 units for Intelligence.Speed#12 / 114125.2Output tokens per second4 out of 4 units for Speed.Cost#12 / 114In $0.435Out $0.87Cache Discount 99%$0.13Cost per Intelligence Index task2 out of 4 units for Cost.Verbosity#18 / 114140MOutput tokens from Intelligence Index3 out of 4 units for Verbosity.Comparison SummaryMiMo-V2.6-Pro is amongst the leading models in intelligence and reasonably priced when comparing to other open weight models of similar size. It's also notably fast, however somewhat verbose. The model supports text, image, speech, and video input, outputs text, and has a 1M tokens context window.MiMo-V2.6-Pro scores 46 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models (median: 18). When evaluating the Intelligence Index, it generated 140M tokens, which is somewhat verbose in comparison to the median of 140M.Pricing for MiMo-V2.6-Pro is $0.43 per 1M input tokens (somewhat expensive, median: $0.30) and $0.87 per 1M output tokens (moderately priced, median: $1.13). In total, it cost $206.66 to evaluate MiMo-V2.6-Pro on the Intelligence Index.At 125 tokens per second, MiMo-V2.6-Pro is notably fast (78).Technical specificationsReasoningYesThis page shows the reasoning version of this model.A non-reasoning variant may also exist.Input modalitySupports: text, image, speech, and videoOutput modalitySupports: textContext window1M~1500 A4 pages of size 12 Arial fontTotal parameters1.0TActive parameters42BNumber of parameters active per token during inferenceLicenseMITModel weightsHugging Face114 models in this classMetrics are compared against models of the same class:Non-reasoning models → compared only with other non-reasoning modelsReasoning models → compared across both reasoning and non-reasoningOpen weights models → compared only with other open weights models of the same size class:Tiny: ≤4B parametersSmall: 4B–40B parametersMedium: 40B–150B parametersLarge: >150B parametersProprietary models → compared across proprietary and open weights models of the same price range, using a blended 3:1 input/output price ratio:<$0.15 per 1M tokens$0.15–$1 per 1M tokens>$1 per 1M tokensModel ComparisonAPI Provider BenchmarksHighlightsUpdatedIntelligenceArtificial Analysis Intelligence Index · Higher is betterSpeedOutput tokens per second · Higher is betterCost per TaskWeighted average cost (USD) per Intelligence Index task · Lower is betterIntelligenceUpdatedCapability IndexesBenchmarksOpenness IndexIntelligence Index ComparisonsToken UseCostContext WindowSpeedLatencyEnd-to-End Response TimeModel Size (Open Weights Models Only)Prompt OptionsIntelligenceUpdatedArtificial Analysis Intelligence IndexArtificial Analysis Intelligence Index v4.3.2 incorporates 10 evaluations: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.128 of 656 modelsAdd model from specific providerArtificial Analysis Intelligence IndexArtificial Analysis Intelligence Index v4.3.2 includes: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.Open Weights / ProprietaryReasoning / Non-ReasoningText Only / Multimodal InputsArtificial Analysis Intelligence Index by Open Weights / ProprietaryArtificial Analysis Intelligence Index v4.3.2 incorporates 10 evaluations: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.128 of 656 modelsAdd model from specific providerProprietaryOpen WeightsOpen Weights (Commercial Use Restricted)Artificial Analysis Intelligence IndexArtificial Analysis Intelligence Index v4.3.2 includes: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.Open WeightsIndicates whether the model weights are available. Models are labelled as 'Commercial Use Restricted' if commercial use is limited by conditions, and as 'Non-commercial' if the license prohibits commercial use.Capability IndexesMeasures the performance of models on specific capabilities and industriesFinance & AccountingStrategy & OpsLegalEngineeringEconomicsArtificial Analysis Finance & Accounting IndexIncorporates 7 evaluations: AA-Omniscience, GDPval-AA v2.1, AA-Briefcase v1.1, Humanity's Last Exam, AutomationBench-AA, AA-LCR v1.1, GDP.pdf · Higher is better28 of 153 modelsAdd model from specific providerBenchmarksIntelligence EvaluationsIntelligence evaluations measured independently by Artificial Analysis · Higher is betterCodingAgenticTool UsePrivate DatasetUser InteractionFinanceMedicalLegalIntelligence IndexLong ContextMultimodalInstruction FollowingFaithfulnessWritingBusinessSee more18 of 26 evaluations28 of 656 modelsAdd model from specific providerAA-Briefcase v1.1UpdatedAgentic knowledge work, (Elo-500)/2000GDPval-AA v2.1UpdatedAgentic real-world work tasks, (Elo-500)/2000AutomationBench-AAUpdatedAgentic SaaS workflowsTerminal-Bench 4.0NewAgentic coding & terminal useSciCodeCodingHumanity's Last ExamReasoning & knowledgeGDP.pdfNewProfessional document reasoning, All-passCritPtPhysics reasoningAA-Omniscience AccuracyKnowledgeAA-Omniscience Non-Hallucination Rate1 - hallucination rateAA-LCR v1.1Long context reasoningHarvey LAB-AALegal agentic work, criterion pass rateEnterpriseOps-Gym-AAAgentic business operationsAA-AnalystAgentQuantitative analysis on spreadsheets & documents𝜏³-BankingAgentic tool useITBench-AAKubernetes incident root-cause analysisMMMU-ProVisual reasoningMLCR-AANewMedical long context reasoningIntelligence Evaluation RelevanceWhile model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.Artificial Analysis Intelligence IndexArtificial Analysis Intelligence Index v4.3.2 includes: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.AA-Briefcase v1.1UpdatedAA-Briefcase EloAA-Briefcase Rubric Score (%)Analytical Quality & Presentation EloAA-Briefcase EloAA-Briefcase v1.1 is an agentic knowledge work benchmark developed by Artificial Analysis. AA-Briefcase Elo is a combined metric that aggregates rubric pass rate, analytical quality Elo and presentation Elo · Higher is better28 of 174 modelsAdd model from specific providerAA-Briefcase EloAA-Briefcase Elo is a combined metric that aggregates analytical quality Elo, presentation Elo, and rubric pass rate, with rubric performance converted into Elo via synthetic head-to-head matches. Elo and 95% confidence interval bounds are clamped at 0.AA-OmniscienceAA-Omniscience IndexAA-Omniscience AccuracyAA-Omniscience Hallucination RateAA-Omniscience IndexAA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.28 of 531 modelsAdd model from specific providerAA-Omniscience IndexAA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.Openness IndexOpenness IndexOpenness Index ComponentsOpenness vs. IntelligenceArtificial Analysis Openness Index: ScoreOpenness Index assesses model openness on a 0 to 100 normalized scale (higher is more open)19 of 322 modelsAdd model from specific providerIntelligence Index ComparisonsIntelligence Index vs. Cost per TaskIntelligence Index vs. Time per TaskIntelligence Index vs. Output SpeedIntelligence Index vs. End-to-End Response TimeIntelligence Index vs. Cost per Intelligence Index TaskArtificial Analysis Intelligence Index · Weighted average cost (USD) per Artificial Analysis Intelligence Index task28 of 656 modelsMost attractive quadrantPareto lineXiaomiOpenAIAnthropicSpaceXAIGoogleMetaZ AIDeepSeekKimiAlibabaMiniMaxTencentThinking MachinesNVIDIACost per Intelligence Index TaskWeighted average cost per Intelligence Index task. Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight.Artificial Analysis Intelligence IndexArtificial Analysis Intelligence Index v4.3.2 includes: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience, AA-LCR v1.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.Token UseOutput Tokens per TaskIntelligence Index vs. Output Tokens per TaskIntelligence Index Token UseIntelligence Index vs. Token UseOutput Tokens per Intelligence Index TaskWeighted average number of output tokens used to run one task in the Artificial Analysis Intelligence Index28 of 656 modelsAnswerReasoningOutput Tokens per Intelligence Index TaskThe number of tokens required per Intelligence Index task. This is calculated by multiplying the output tokens per eval by the relative weights of each benchmark in the Intelligence Index, then dividing by task count (excluding repeats).CostCost per TaskIntelligence Index vs. Cost per TaskEvaluation BreakdownCost per Intelligence Index TaskWeighted average cost (USD) per Artificial Analysis Intelligence Index task, segmented by token type. Lower is better28 of 656 modelsAnswerReasoningCache WriteCache HitInputCost per Intelligence Index TaskWeighted average cost per Intelligence Index task. Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight.Intelligence Index Total CostIntelligence Index vs. Total CostCost to Run Artificial Analysis Intelligence IndexCost (USD) to run all evaluations in the Artificial Analysis Intelligence Index28 of 656 modelsAdd model from specific providerOutputReasoningCache WriteCache ReadNon-Cache InputCost to Run Artificial Analysis Intelligence IndexThe cost to run the evaluations in the Artificial Analysis Intelligence Index, calculated using the model's input, cache hit, cache write, reasoning, and answer token prices and the number of tokens used across evaluations (excluding repeats).Cache Hit, Input, and Output PricingBlended PriceBlended Price (Stacked)Cache DiscountIntelligence Index vs. PriceIntelligence Index vs. Price (Log, Inverted)Image Input PricingPricing: Cache Hit, Input, and OutputPrice (USD per M Tokens)28 of 656 modelsAdd model from specific providerCache HitInputOutputCache HitPrice per token for cached prompts (previously processed), typically offering a significant discount compared to regular input price, represented as USD per million tokens. The values shown here are the cache hit price; cache write and cache storage are billed separately and vary by provider — see "Cache pricing by provider" for detail. 4 more notesContext WindowContext WindowIntelligence Index vs. Context WindowContext WindowContext window: tokens limit · Higher is better28 of 656 modelsAdd model from specific providerContext Window for RAGLarger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.Context WindowMaximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).SpeedMeasured by Output Speed (tokens per second)Output SpeedOutput Speed by Prompt TypeOutput Speed VarianceOutput Speed Over TimeOutput Speed vs. PriceLatency vs. Output SpeedOutput SpeedOutput tokens per second · Higher is better28 of 656 modelsAdd model from specific providerOutput SpeedTokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).Model Performance RepresentationFigures represent performance of the model's first-party API or the median across providers where a first-party API is not available.Time per TaskIntelligence Index vs. Time per TaskCost vs. Time per TaskTime per Intelligence Index TaskWeighted average decode time (minutes) per task; excludes TTFT and overhead time · Lower is better28 of 656 modelsTime per Intelligence Index TaskThe weighted average time (seconds) per Artificial Analysis Intelligence Index task. This is calculated by dividing output tokens per task by output speed, weighted by the relative weights of each benchmark in the Intelligence Index.LatencyMeasured by Time (seconds) to First TokenTime To First Answer TokenTime To First TokenLatency by Prompt TypeLatency VarianceLatency Over TimeLatency: Time To First Answer TokenSeconds to first answer token received · Accounts for reasoning model 'thinking' time28 of 656 modelsAdd model from specific providerThinking (reasoning models, when applicable)Input processingTime to First Answer TokenTime to first answer token received, in seconds, after API request sent. For reasoning models, this includes the 'thinking' time of the model before providing an answer. For models which do not support streaming, this represents time to receive the completion.End-to-End Response TimeSeconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speedEnd-to-End Response TimeEnd-to-End Response Time by Prompt TypeEnd-to-End Response Time Over TimeEnd-to-End Response TimeSeconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better28 of 656 modelsAdd model from specific providerOutputting time'Thinking' time (reasoning models)Input processing timeEnd-to-End Response TimeSeconds to receive a 500 token response. Key components:Input time: Time to receive the first response tokenThinking time (only for reasoning models): Time reasoning models spend outputting tokens to reason prior to providing an answer. Amount of tokens based on the average reasoning tokens across a diverse set of 60 prompts (methodology details).Answer time: Time to generate 500 output tokens, based on output speedModel Performance RepresentationFigures represent performance of the model's first-party API or the median across providers where a first-party API is not available.Model Size (Open Weights Models Only)Total & Active ParametersIntelligence Index vs. Active ParametersIntelligence Index vs. Total ParametersModel Size: Total and Active ParametersComparison between total model parameters and parameters active during inference28 of 656 modelsAdd model from specific providerActive ParametersPassive ParametersTotal ParametersThe total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.Active Parameters at Inference TimeThe number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.Frequently Asked QuestionsCommon questions about MiMo-V2.6-ProWhen was MiMo-V2.6-Pro released?MiMo-V2.6-Pro was released on September 21, 2026.Who created MiMo-V2.6-Pro?MiMo-V2.6-Pro was created by Xiaomi.How intelligent is MiMo-V2.6-Pro?MiMo-V2.6-Pro scores 46 on the Artificial Analysis Intelligence Index, placing it well above average among other open weight models of similar size (median: 18).How fast is MiMo-V2.6-Pro?MiMo-V2.6-Pro generates output at 125.2 tokens per second (based on Xiaomi's API), which is well above average compared to other open weight models of similar size (median: 77.7 t/s).What is the latency of MiMo-V2.6-Pro?MiMo-V2.6-Pro has a time to first token (TTFT) of 2.19s (based on Xiaomi's API), which is better than average compared to other open weight models of similar size (median: 2.27s).How much does MiMo-V2.6-Pro cost?MiMo-V2.6-Pro costs $0.43 per 1M input tokens (better than average, median: $0.45) and $0.87 per 1M output tokens (very competitive, median: $1.68), based on Xiaomi's API.What is MiMo-V2.6-Pro API pricing?MiMo-V2.6-Pro costs $0.43 per 1M input tokens and $0.87 per 1M output tokens (based on Xiaomi's API). For a blended rate (7:2:1 cache hit/input/output ratio), this is $0.18 per 1M tokens. Pricing may vary by provider. Compare provider pricingHow verbose is MiMo-V2.6-Pro?When evaluated on the Intelligence Index, MiMo-V2.6-Pro generated 140M output tokens, which is somewhat higher than average compared to other open weight models of similar size (median: 140M).Is MiMo-V2.6-Pro a reasoning model?Yes, MiMo-V2.6-Pro is a reasoning model. It uses extended thinking or chain-of-thought reasoning to work through complex problems before providing an answer.What input modalities does MiMo-V2.6-Pro support?MiMo-V2.6-Pro supports text, image, speech, and video input.What output modalities does MiMo-V2.6-Pro support?MiMo-V2.6-Pro supports text output.Can MiMo-V2.6-Pro process images?Yes, MiMo-V2.6-Pro supports image input and can analyze, describe, and answer questions about images.Is MiMo-V2.6-Pro multimodal?Yes, MiMo-V2.6-Pro is multimodal. It can process text, image, speech, and video input and generate text output.What is the context window of MiMo-V2.6-Pro?MiMo-V2.6-Pro has a context window of 1.0M tokens. This determines how much text and conversation history the model can process in a single request.Is MiMo-V2.6-Pro open source?Yes, MiMo-V2.6-Pro is open weights. The model weights are publicly available and can be downloaded for self-hosting.How many parameters does MiMo-V2.6-Pro have?MiMo-V2.6-Pro has 1.0 trillion parameters (42 billion active).What are the active parameters of MiMo-V2.6-Pro?MiMo-V2.6-Pro is a Mixture of Experts (MoE) model with 1.0 trillion total parameters, but only 42 billion active parameters are used during inference.What is the license for MiMo-V2.6-Pro?MiMo-V2.6-Pro is released under the MIT license. This license allows commercial use.How does MiMo-V2.6-Pro perform on benchmarks?MiMo-V2.6-Pro achieves a score of 46 on the Artificial Analysis Intelligence Index. This composite benchmark evaluates models across reasoning, knowledge, mathematics, and coding.Is MiMo-V2.6-Pro available via API?Yes, MiMo-V2.6-Pro is available via API through 1 provider. Compare API providersWhere can I use MiMo-V2.6-Pro?MiMo-V2.6-Pro is available through 1 API provider. Compare providersArtificial AnalysisGet notified about new articlesEmail addressSubscribeArtificial AnalysisExploreLLM LeaderboardImage ArenaVideo ArenaAI AgentsEvaluationsProductsOptimaMicroEvalsModel RecommenderData PlaygroundImage LabCompanyAboutMethodologyContactArticlesXLinkedInYouTubeRednoteDiscord© 2026 Artificial AnalysisTerms of UseData Platform TermsPrivacy PolicyEnglish

MiMo-V2.6-Pro is presented as a leading open weights model, released in September 2026 by Xiaomi, which demonstrates strong performance when analyzed across intelligence, speed, and cost relative to other models of similar scale. The model achieved a score of 46 on the Artificial Analysis Intelligence Index, positioning it favorably above the median score of 18 among comparable models, suggesting robust reasoning and knowledge capabilities. Performance benchmarks indicate that MiMo-V2.6-Pro operates at a speed of 125.2 output tokens per second, which is notably fast compared to the median output speed for models in this class.

The model possesses significant multimodal capabilities, supporting the processing of text, image, speech, and video inputs, and generating text outputs. Furthermore, it features a substantial context window of 1.0 million tokens, enabling it to handle extensive context necessary for complex reasoning tasks. Architecturally, MiMo-V2.6-Pro is a large model, boasting a total of 1.0 trillion parameters, although only 42 billion of these parameters are active during inference, indicating that it operates as a Mixture of Experts (MoE) model. This structure is reflected in its reasoning capacity, as the model utilizes extended thinking or chain-of-thought reasoning to tackle intricate problems before formulating responses.

In terms of operational economics, the pricing structure for MiMo-V2.6-Pro is $0.43 per 1 million input tokens and $0.87 per 1 million output tokens. While the input cost is competitive, the output cost is moderately priced relative to the median, reflecting a blended rate that is highly attractive when considering cache hit discounts. The overall evaluation process cost for the Intelligence Index task amounted to $206.66.

When compared to the broader landscape of artificial analysis benchmarks, the model's verbosity is slightly high, generating 140 million output tokens during the Intelligence Index evaluation, slightly exceeding the median, which warrants consideration in task-specific deployment strategies. The model is released under the permissive MIT license, which permits commercial use, further enhancing its utility for practical applications. The extensive benchmark suite, which includes evaluations focused on agentic knowledge work, coding, scientific knowledge, and general intelligence, suggests a broad and versatile capability assessment across various domains, including finance, legal, engineering, and mathematics.