Skip to content

Xiaomi mit Released Apr 22, 2026 Updated

MiMo-V2.5

ToolsGuide Scores, rankings, test results and price.

Model summary

  • Text #103 / 416

    370 / 1,000

    ToolsGuide Score· votes only

  • Vision #75 / 167

    389 / 1,000

    ToolsGuide Score· votes 397, tests 380

  • Coding #77 / 147

    308 / 1,000

    ToolsGuide Score· votes 311, tests 306

  • Price per 1M tokens

    In $0.17· Out $0.34

    $0.17

    Input · median text model $0.43

Summary

MiMo-V2.5 is Xiaomi's model, released Apr 22, 2026. It's ranked in 3 categories and does best at text: #103 of 416, with a ToolsGuide Score of 370 against a median of 235.

It also ranks #75 of 167 for vision (389), #77 of 147 for coding (308).

Its text score uses votes only, because it has published fewer than two of the tests used for that category.

It costs $0.17 per million input tokens and $0.34 per million output tokens, below the median text model ($0.43).

It has published results on 14 tests and has the top score on 2 of them, including HR-Bench (4k) and ResearchClawBench.

Specifications

Creator
Xiaomi
Licence
mit
Released
Apr 22, 2026
Context window
1M tokens
Input price
$0.17 / 1M
Output price
$0.34 / 1M
Tests reported
14
Settings ranked
mimo-v2.5, MiMo-V2.5

Model comparison

Against the ten best text models.

Text score

ToolsGuide Score out of 1,000 · Higher is better

Vision score

ToolsGuide Score out of 1,000 · Higher is better

Input price

USD per 1M input tokens · Lower is better

Leaderboards

Where MiMo-V2.5 sits among the top 20 in each category.

Text

#103 of 416 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model

Full text leaderboard →

Vision

#75 of 167 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model

Full vision leaderboard →

Coding

#77 of 147 · ToolsGuide Score out of 1,000, where 1,000 is a flawless model

Full coding leaderboard →

By task

Scored from votes on each kind of prompt alone, on the same 1,000-point scale.

Score against price

Text · the 10 priced models ranked closest · up and to the left is better value

Against its nearest rivals

Text prompts by kind · ToolsGuide Score from votes · hover for values

Text

mimo-v2.5 · hover a bar for its rank

Vision

MiMo-V2.5 · hover a bar for its rank

Coding

MiMo-V2.5 · hover a bar for its rank

Test results

Every test MiMo-V2.5 has a published result on, out of 100%, with its rank among the models that took it.

  • MMMU-Pro

    #23 / 73

    College-level questions that need both an image and text to answer.

    77.9%

    Self-reported · In the vision score

  • SWE-Bench Pro

    #39 / 61

    Harder, longer software engineering tasks from real codebases.

    56.1%

    Self-reported · In the coding score

  • CharXiv-R

    #26 / 60

    Reading and reasoning about charts from scientific papers.

    81.0%

    Self-reported

  • Terminal-Bench 2.0

    #16 / 54

    Getting real work done in a command-line terminal.

    65.8%

    Self-reported

  • Finance Agent v2

    #23 / 28

    Finance Agent v2 is an agentic financial-analysis benchmark from Vals that evaluates models on real-world finance workflows, measuring their ability to retrieve and reason over financial documents, perform multi-step calculations, and produce accurate analyses.

    36.7%

    Self-reported

  • Video-MME

    #4 / 17

    Video-MME is the first-ever comprehensive evaluation benchmark of Multi-modal Large Language Models (MLLMs) in video analysis. It features 900 videos totaling 254 hours with 2,700 human-annotated question-answer pairs across 6 primary visual domains (Knowledge, Film & Television, Sports Competition, Life Record, Multilingual, and others) and 30 subfields. The benchmark evaluates models across diverse temporal dimensions (11 seconds to 1 hour), integrates multi-modal inputs including video frames, subtitles, and audio, and uses rigorous manual labeling by expert annotators for precise assessment.

    87.7%

    Self-reported

  • Claw-Eval

    #7 / 14

    Claw-Eval tests real-world agentic task completion across complex multi-step scenarios, evaluating a model's ability to use tools, navigate environments, and complete end-to-end tasks autonomously.

    63.2%

    Self-reported

  • MiMo Coding Bench

    #2 / 4

    MiMo Coding Bench evaluates coding-agent capabilities on software engineering tasks reported with the MiMo model family.

    71.8%

    Self-reported

  • VideoHolmes

    #3 / 4

    VideoHolmes evaluates video understanding and reasoning capabilities in multimodal models.

    64.0%

    Self-reported

  • GraphWalks

    #2 / 3

    GraphWalks is a synthetic multi-hop long-context reasoning benchmark in which a model is given an edge-list representation of a graph and must traverse it to find neighboring nodes (via breadth-first search) or parent nodes for a given start node. Performance is reported as F1 of the model-predicted answer set versus the ground truth.

    87.0%

    Self-reported

  • DailyOmni

    #2 / 2

    DailyOmni evaluates multimodal models on daily-life video understanding tasks.

    83.5%

    Unverified

  • OmniDocBench

    #2 / 2

    OmniDocBench evaluates multimodal models on document understanding tasks such as OCR, layout parsing, and structured document comprehension.

    87.2%

    Unverified

  • HR-Bench (4k)

    #1 / 1

    HR-Bench (4k) evaluates image understanding on high-resolution visual inputs with a 4k setting.

    88.5%

    Unverified

  • ResearchClawBench

    #1 / 1

    ResearchClawBench evaluates research agents on realistic, tool-using research tasks that require code execution and filesystem workspace interaction.

    16.9%

    Unverified