Inference
EN
Xiaomi

MiMo-V2.6-Flash

MiMo-V2.6-Flash is a reasoning model from Xiaomi in the MiMo-V2.6 family. 23 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text · image · video · audio
  • Released 2026-09
RankedEnough results for an index.
IndexAll categories, out of 100.64.3±4.4Median 51.0
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.28/sMedian 67/s
First tokenSeconds to the first output token.3.3 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.14Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.28Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1450±5 #54Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
xiaomi/mimo-v2.6-flash
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$0.10$0.251.05M
Deep InfraHost$0.14$0.281.05M
Novita AIHost$0.14$0.281.05M
OpenRouterRouter$0.14$0.281.05M
XiaomiMaker$0.14$0.281.05M

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

Run locally

310.76B parameters
16-bit
653.6 GB
8-bit
347.7 GB
4-bit
198.8 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.12 counted69.726 / 218
CodingCode writing and repair.5 counted64.149 / 209
ReasoningLogic problems and puzzles.2 counted59.269 / 227
MultimodalTasks with images and text.1 counted50.3—
KnowledgeFacts and expert knowledge.2 counted52.3118 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.None counted——
MathMath problems.1 counted68.9—

Results

23 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
CyberGymAgentic95.1%1 / 29
GDPval-AA normalizedAgentic55.4%19 / 133
Terminal-Bench 2.1 · BenchLMAgentic87.6%13 / 86
Terminal-Bench 2.1 · Vals AIAgentic76.4% ±1.716 / 75
Critical Physics TasksReasoning12.0%57 / 214
Vibe Code Bench v1.1Coding79.0% ±4.033 / 123
OSWorld-VerifiedAgentic80.8%10 / 37
JobBenchAgentic61.2%10 / 36
Artificial Analysis Humanity's Last ExamKnowledge35.1%62 / 222
AutomationBenchAgenticAbove fitFar above this model's index.52.3%9 / 32
Agent Arena task outcomeAgentic4.219 / 54
Agent Arena steerabilityAgentic2.820 / 54
Code MigrationCoding40.9% ±4.429 / 77
Artificial Analysis SciCodeCoding51.3%46 / 114
ProofBench v1.1Math63.0% ±4.923 / 54
DeepSWEAgentic67.9%26 / 59
Artificial Analysis Long Context ReasoningReasoning74.3%98 / 217
Toolathlon-VerifiedAgentic73.6%11 / 23
Terminal-Bench 4.0Agentic24.2% ±1.530 / 61
Artificial Analysis Omniscience AccuracyKnowledge27.2%91 / 164
Artificial Analysis MMMU-ProMultimodal73.1%66 / 113
IOICoding47.7% ±4.827 / 45
Agent Arena command recoveryAgentic−1.136 / 50
Agents' Last ExamAgentic27.6%16 / 21
ProgramBench: Can Language Models Rebuild Programs From Scratch?Coding26.0%16 / 19
ExploitGymAgentic6.0%21 / 23
CyberBench v1.1Other75.4% ±5.47 / 48
Finance Agent (v2)Finance56.3% ±0.519 / 79
MedScribeHealthcare85.3% ±2.029 / 109
Public Benefits Bench v1.1Public benefits67.6% ±1.214 / 50
ProgramBenchCoding0.5% ±0.519 / 65
EMBFinance65.5% ±2.724 / 75
HLAB-Vals-1.1Legal13.3% ±2.814 / 35
Tax Agent BenchFinance28.5% ±3.137 / 82
Legal Research BenchLegal38.0% ±3.443 / 90
BioMysteryBenchScience69.3% ±1.015 / 30
SRE BenchOther4.2% ±1.218 / 36
MedCodeHealthcare41.1% ±2.058 / 106
Terminal-Bench-ScienceScience4.3% ±2.426 / 44
SAGEEducation83.3% ±2.122 / 29
MysteryMechanismScience21.6% ±2.823 / 30

Sources

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

History

5 releases
Index64.9 → 64.3
23 Oct 202612 Nov 2026
Output / 1M$0.28 → $0.28
23 Oct 202612 Nov 2026

Questions

How much does MiMo-V2.6-Flash cost?
$0.14 per 1M input tokens, $0.28 per 1M output tokens.
How good is MiMo-V2.6-Flash?
Its index is 64.3 out of 100. That is place 49 of 283 ranked models.
What is MiMo-V2.6-Flash best at?
Agentic: 69.7 out of 100, place 26 of 218.
How fast is MiMo-V2.6-Flash?
28 output tokens per second, 3.3 s to the first token.
What is the context window of MiMo-V2.6-Flash?
1.05M tokens.
Is MiMo-V2.6-Flash open source?
Yes. Its weights are published.
When was MiMo-V2.6-Flash released?
September 2026.

Badge

MiMo-V2.6-Flash on Inference360: place 49 of 283, index 64.3
HTML
<a href="https://inference360.ai/models/mimo-v2-6-flash"><img src="https://inference360.ai/badges/mimo-v2-6-flash.svg" alt="MiMo-V2.6-Flash on Inference360: place 49 of 283, index 64.3" height="20"></a>
Markdown
[![MiMo-V2.6-Flash on Inference360: place 49 of 283, index 64.3](https://inference360.ai/badges/mimo-v2-6-flash.svg)](https://inference360.ai/models/mimo-v2-6-flash)

Compare with

Qwen3.8-Flash-Next64.3Claude Sonnet 564.2Gemini 3.5 Flash64.2Muse Spark 1.164.6MiMo-V2.6-Pro70.2

Lower price

dots3-note Preview63.4 · FreeGLM-5.3-Flash63.1 · $0.078/1MLing 3.1 Flash67.2 · $0.135/1M

More from Xiaomi

All models
MiMo-V2.6-Pro70.2MiMo V2.6 Pro UltraSpeedUnrankedMiMo-V2.5-Pro56.3MiMo-V2.554.8MiMo-V2-Pro51.8