BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases16
XiaomiOpen weights: XiaomiMiMo/MiMo-V2.6-Pro-RLReasoning modelCompare with another model

MiMo-V2.6-Pro

MiMo-V2.6-Pro is a Xiaomi open-weights reasoning model, released 21 Sept 2026. Its best configuration ranks #54 of 775 on the BenchLeader Index at 64.2 ±4.2, in the upper half. The ± is the point: 134 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in reasoning (79) and lowest in agents & tools (59). At $0.548 per million tokens blended it is mid-priced. Output speed of 50 tokens per second puts it slower than most, with a first answer in 43.1 s. Last measured 19 Oct 2026.

Blended price
$0.548/M
$0.440 in · $0.870 out
Output speed
50 tok/s
measured by Artificial Analysis
First answer
43 s
first token 2.97 s
Context
1.0M
Full answer
53 s
median, reasoning included
Cost per run
$0.133
one full Intelligence Index run
Released
21 Sept 2026
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Overall index64
  2. Reasoning79
  3. Coding62
  4. Agents & tools59
  5. Maths66
  6. Knowledge65
  7. Human preference67
  8. Multimodal60
  9. Long context68
  10. Composite70

Versions

Xiaomi has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

ModelReleasedIndexRank
MiMo-V2.6-Prothis page21 Sept 202664.2#54
MiMo-V2.5-Prothinking23 Apr 202657.0#190

How its index has moved

23 Sept 2026 to 19 Oct 2026
576470

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. 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.

Metadata

Benchmarknot statedSourceTrend
AA Openness Indexnot in index52.8#26Artificial Analysis

Reasoning

Benchmarknot statedSourceTrend
Arena Hard Prompts1504#25Arena (LMArena)
LiveBench 2.0 Reasoningnot in index79.6%#40LiveBench
Humanity's Last Exam (AA)not in index49.4%#28Artificial Analysis
CritPt26.6%#37Artificial Analysis
MysteryMechanismnot in index15.3%#25Vals AI

Coding

Benchmarknot statedSourceTrend
SciCode60.9%#7SciCode
Arena Coding1538#13Arena (LMArena)
Arena WebDev1625#20Arena (LMArena)
LiveBench 2.0 Codingnot in index81.5%#29LiveBench
SciCode (AA)not in index60.9%#8Artificial Analysis
IOI39.3%#35Vals AI
Code Migrationnot in index43.0%#26Vals AI
ProgramBenchnot in index0.5%#18Vals AI
Vibe Code Bench v1.1not in index85.2%#15Vals AI
ALE-Benchnot in index1158.1#48Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index34.9%#34Artificial Analysis
FrontierCode 1.1 (AA)not in index45.4%#11Artificial Analysis

Agents & tools

Benchmarknot statedSourceTrend
APEX-Agents54.4%#22Mercor
Arena Agent3.7#13Arena (LMArena)
LiveBench Long Agenticnot in index47.5%#18LiveBench
LiveBench Browsernot in index88.0%#7LiveBench
Terminal-Bench 2.1 (Vals)67.8%#32Vals AI
GDPval-AA v2.1not in index59.2%#15Artificial Analysis
Analyst Agent (AA)not in index37.5%#26Artificial Analysis
CyberBenchnot in index72.9%#14Vals AI
Finance Agent v2not in index57.3%#15Vals AI
Harvey's Legal Agent Benchmarknot in index13.8%#8Vals AI
Legal Research Benchnot in index47.1%#14Vals AI
SREBenchnot in index3.0%#22Vals AI
Tax Agent Benchnot in index32.4%#17Vals AI
Terminal-Bench 4.0 (Vals)not in index31.3%#19Vals AI
Terminal-Bench Sciencenot in index2.9%#32Vals AI
AutomationBenchnot in index58.6%#42Artificial Analysis
GDP.pdf (AA)not in index19.2%#68Artificial Analysis
AA-Briefcase v1.1not in index1515#27Artificial Analysis
Terminal-Bench-Science 0.1 (AA)not in index5.7%#35Artificial Analysis
CWE-Bench (AA)63.3%#4Artificial Analysis
DeepSec-Bench (AA)26.2%#6Artificial Analysis
CyberGym end-to-end (AA)78.6%#2Artificial Analysis

Maths

Benchmarknot statedSourceTrend
ProofBench70.0%#16Vals AI
Arena Maths1488#29Arena (LMArena)

Knowledge

Benchmarknot statedSourceTrend
LiveBench 2.0 Data Analysisnot in index72.3%#32LiveBench
LiveBench Researchnot in index76.7%#21LiveBench
AA-Omniscience8.4#97Artificial Analysis
Excel Modeling Benchmarknot in index62.9%#29Vals AI
MedCodenot in index45.0%#39Vals AI
MedScribenot in index88.3%#13Vals AI
Public Benefits Benchnot in index68.9%#8Vals AI
AA-Omniscience: accuracynot in index34.9%#152Artificial Analysis
AA-Omniscience: non-hallucinationnot in index59.4%#81Artificial Analysis

Instruction following

Benchmarknot statedSourceTrend
Arena Instruction Followingnot in index1471#35Arena (LMArena)

Human preference

Benchmarknot statedSourceTrend
Arena Text1479#27Arena (LMArena)
Arena Creative Writingnot in index1445#52Arena (LMArena)
Arena Multi-turnnot in index1455#80Arena (LMArena)
Arena Longer Queriesnot in index1494#22Arena (LMArena)

Multimodal

Benchmarknot statedSourceTrend
Arena Vision1263#53Arena (LMArena)
LiveBench Designnot in index78.5%#18LiveBench
SAGEnot in index45.0%#41Vals AI
AA-3Dnot in index980#13Artificial Analysis

Long context

Benchmarknot statedSourceTrend
AA-LCR86.3%#3Artificial Analysis
MLCRnot in index18.3%#33Artificial Analysis

Composite

Benchmarknot statedSourceTrend
LiveBench 2.074.9%#18LiveBench
AA Intelligence Index v4.3.246.3#33Artificial Analysis
Vals Indexnot in index57.8#12Vals AI
AA Cyber Indexnot in index56.1#2Artificial Analysis

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

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
DeepInfra37 tok/s1.76 s$0.387$0.7831.0Mfp8
NovitaAI35 tok/s3.46 s$0.435$0.8701.0Mfp8
GMICloud29 tok/s4.42 s$0.435$0.8701.1Mbf16
Xiaomi24 tok/s2.97 s$0.435$0.8701.0Mfp8

Price history

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

$0.00$0.239$0.479$0.718$0.957Sept 26
input outputnow $0.435 in · $0.870 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 $0.004 per 1M. Reasoning tokens are not modelled.

WorkloadTokens in / outCostWith cachingTime
Chat reply400 / 300$0.0004$0.000349.1 s
Summarize a 30-page report12,000 / 600$0.0058$0.001955.1 s
Code edit6,000 / 1,500$0.0039$0.00201.2 min
Agentic coding session60,000 / 4,000$0.030$0.0102.0 min
Structured extraction2,000 / 200$0.0011$0.000447.1 s

More from Xiaomi

New this month

Compare MiMo-V2.6-Pro with…

See also

Best cheap LLMBest open-weights LLMBest under $2/MBest open weightsCheapest 1M contextBest for codingAll Xiaomi models

Data as of 21 Oct 2026. Compare with another model.

Cite as: BenchLeader, “MiMo-V2.6-Pro: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/mimo-v2-6-pro, data as of 21 Oct 2026.