Inference
EN
Xiaomi

MiMo-V2.6-Pro

MiMo-V2.6-Pro is a reasoning model from Xiaomi in the MiMo-V2.6 family. 36 benchmarks count toward its index, in 5 categories.

  • Open weights
  • Reasoning
  • Input: text · image · video · audio
  • Released 2026-09
RankedEnough results for an index.
IndexAll categories, out of 100.70.2±4.7Median 51.0
CoverageIndex weight that has results.80%
SpeedOutput tokens per second.31/sMedian 67/s
First tokenSeconds to the first output token.3.9 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.435Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.87Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1492±9 #11Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
xiaomi/mimo-v2.6-pro
Max outputMaximum output tokens per request.
131K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
<$0.01
WeightsLink to the published weights.
Hugging Face (opens in a new tab) ↗
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
5 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
VultrCloud$0.40$0.801.05M
Deep InfraHost$0.43$0.871.05M
Novita AIHost$0.435$0.871.05M
OpenRouterRouter$0.435$0.871.05M
XiaomiMaker$0.435$0.871.05M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Run locally

1024.22B parameters
16-bit
2151.9 GB
8-bit
1143.6 GB
4-bit
653.0 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy needs more than 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.22 counted72.115 / 218
CodingCode writing and repair.7 counted66.633 / 209
ReasoningLogic problems and puzzles.4 counted68.328 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.2 counted64.247 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.None counted——
MathMath problems.1 counted71.3—

Results

36 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Artificial Analysis Long Context ReasoningReasoning86.3%3 / 217
Terminal-Bench 2.1 · BenchLMAgentic89.9%4 / 86
Artificial Analysis Humanity's Last ExamKnowledge49.4%14 / 222
GDPval-AA normalizedAgentic59.2%9 / 133
CyberGymAgentic94.0%2 / 29
Artificial Analysis SciCodeCoding60.9%8 / 114
Critical Physics TasksReasoningAbove fitFar above this model's index.26.6%19 / 214
Vibe Code Bench v1.1Coding85.2% ±3.416 / 123
Artificial Analysis CyberGym-E2E-AAAgenticAbove fitFar above this model's index.78.6%3 / 19
Artificial Analysis Cyber IndexAgentic56.1%3 / 19
LiveBench Browser UseAgentic88.0%8 / 49
JobBenchAgentic62.0%7 / 36
Artificial Analysis CWE-Bench-AAAgentic63.3%4 / 19
OSWorld-VerifiedAgentic82.0%8 / 37
AutomationBenchAgenticAbove fitFar above this model's index.53.1%7 / 32
Toolathlon-VerifiedAgentic76.9%6 / 23
Agent Arena task outcomeAgentic5.016 / 54
DeepSWEAgentic71.9%18 / 59
LiveBench Long AgenticAgentic47.5%16 / 51
Artificial Analysis DeepsecBench-AAAgentic26.2%7 / 21
Code MigrationCoding43.0% ±4.326 / 77
LiveBench DesignAgentic78.5%18 / 52
ProofBench v1.1Math70.0% ±4.621 / 54
Terminal-Bench 4.0Agentic31.3% ±3.124 / 61
Artificial Analysis Omniscience AccuracyKnowledge34.8%67 / 164
LiveBench ResearchAgentic76.7%22 / 52
Terminal-Bench 2.1 · Vals AIAgentic67.8% ±1.932 / 75
Agent Arena steerabilityAgentic1.525 / 54
Agent Arena command recoveryAgentic2.324 / 50
Agents' Last ExamAgentic31.6%12 / 21
LiveBench Coding 2.0Coding81.5%30 / 52
ExploitGymAgentic17.8%14 / 23
LiveBench Data AnalysisReasoning72.3%49 / 80
Artificial Analysis Terminal-Bench v4.0Agentic34.8%14 / 22
Artificial Analysis AutomationBenchAgentic58.6%15 / 23
Multimodal Knowledge Agent BenchmarkAgentic37.3%13 / 19
Bug Hunt BenchCoding22.7%18 / 25
LiveBench ReasoningReasoning79.6%59 / 80
ProgramBench: Can Language Models Rebuild Programs From Scratch?Coding26.5%15 / 19
IOICoding39.3% ±2.436 / 45
Artificial Analysis GDP.pdfAgentic19.2%18 / 22
Medical Long Context Reasoning (MLCR-AA)Reasoning12.8%21 / 22
Artificial Analysis AnalystAgentAgentic37.5%26 / 26
MedScribeHealthcare88.3% ±1.913 / 109
Public Benefits Bench v1.1Public benefits68.9% ±1.28 / 50
Legal Research BenchLegal47.1% ±3.516 / 90
Finance Agent (v2)Finance57.3% ±0.616 / 79
Tax Agent BenchFinance32.4% ±3.422 / 82
CyberBench v1.1Other72.9% ±5.514 / 48
ProgramBenchCoding0.5% ±0.519 / 65
MedCodeHealthcare45.0% ±2.139 / 106
HLAB-Vals-1.1Legal13.8% ±2.813 / 35
EMBFinance62.9% ±3.130 / 75
SRE BenchOther3.1% ±1.122 / 36
SAGEEducation84.4% ±1.920 / 29
Terminal-Bench-ScienceScience2.9% ±2.033 / 44
MysteryMechanismScience15.3% ±2.428 / 30
STEMBench120: six-domain scientific research progressReasoning2.2%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseLiveBenchNo licence statedVals AINo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index70.0 → 70.2
23 Oct 202612 Nov 2026
Output / 1M$0.87 → $0.87
23 Oct 202612 Nov 2026

Questions

How much does MiMo-V2.6-Pro cost?
$0.435 per 1M input tokens, $0.87 per 1M output tokens.
How good is MiMo-V2.6-Pro?
Its index is 70.2 out of 100. That is place 21 of 283 ranked models.
What is MiMo-V2.6-Pro best at?
Agentic: 72.1 out of 100, place 15 of 218.
How fast is MiMo-V2.6-Pro?
31 output tokens per second, 3.9 s to the first token.
What is the context window of MiMo-V2.6-Pro?
1.05M tokens.
Is MiMo-V2.6-Pro open source?
Yes. Its weights are published.
When was MiMo-V2.6-Pro released?
September 2026.

Badge

MiMo-V2.6-Pro on Inference360: place 21 of 283, index 70.2
HTML
<a href="https://inference360.ai/models/mimo-v2-6-pro"><img src="https://inference360.ai/badges/mimo-v2-6-pro.svg" alt="MiMo-V2.6-Pro on Inference360: place 21 of 283, index 70.2" height="20"></a>
Markdown
[![MiMo-V2.6-Pro on Inference360: place 21 of 283, index 70.2](https://inference360.ai/badges/mimo-v2-6-pro.svg)](https://inference360.ai/models/mimo-v2-6-pro)

Compare with

Muse Spark 1.370.1Claude Mythos 570.8MiniMax M3.1 Flash69.3Grok 4.769.1GPT-5.569.1

Lower price

Ling 3.1 Flash67.2 · $0.135/1MDeepSeek V4.1 Flash 101768.5 · $0.40/1MMiniMax M3.1 Flash69.3 · $0.60/1M

More from Xiaomi

All models
MiMo-V2.6-Flash64.3MiMo V2.6 Pro UltraSpeedUnrankedMiMo-V2.5-Pro56.3MiMo-V2.554.8MiMo-V2-Pro51.8