XiaomiEco TierMoE SparseZero Data Retention

MiMo V2.5 Pro

Name: ARMES
Availability: InStock
Author: ARMES

SWE-Bench Pro: 57.2%, Artificial Analysis Intelligence Index: 54, Claw-Eval Multimodal: 23.8 (matches Sonnet 4.6), Video-MME: 87.7 (rivals Gemini 3 Pro), HLE: 48.0%, autonomous workflows with 1,000+ tool calls | Complex software engineering, long-horizon agentic tasks, real-time video/audio analysis, multimodal chart interpretation, multi-step automated workflows.

Run MiMo V2.5 Pro

Included on ECO plan

Context Window

Input Token Price

$1.00

per 1,000,000 tokens

Output Token Price

$3.00

per 1,000,000 tokens

Speed Rating

Slow

Slow (extended reasoning)

Verified Frontier Benchmarks

Empirical Evaluation Results

Evaluated across standard and verified harnesses

AA Intelligence Index

reasoning

26.0

AA Coding Index

coding

60.2

AnalystAgent

strategy

20.0%

Apex Agents

agentic

2.4%

AutomationBench

agentic

13.7%

AA Briefcase

strategy

883.4 Elo

Critpt

reasoning

4.0%

Gdp Pdf

strategy

4.0%

AA GDPval 2.0

strategy

1186.3 Elo

GPQA Diamond

reasoning

86.6%

Harvey LAB-AA

legal

73.3%

Humanity's Last Exam (HLE)

reasoning

35.7%

Ifbench

reasoning

79.9%

Itbench

reasoning

38.2%

Lcr (1.1)

reasoning

79.7%

Mlcr

reasoning

9.4%

Omniscience

science

3.3

SciCode

science

50.6%

Tau2 Telecom

finance

94.2%

Tau3 Banking

finance

9.9%

Terminal Bench Hard (74221fb)

coding

43.2%

Terminal-Bench 2.1

coding

65.2%

Terminal-Bench 4.0

coding

0.0%

Openrouter GPQA Diamond

reasoning

84.2%

Openrouter Tau2 Bench Airline

finance

74.1%

EQ-Bench

voice

1208.3 Elo

AA Agentic Index

agentic

22.7

TAU-Bench Banking

agentic

9.9%

TAU-2 Bench

agentic

94.2%

Briefcase

strategy

883.4 Elo

Gdpval (2)

strategy

1186.3 Elo

TAU-2 Telecom

agentic

94.2%

TAU-3 Banking

agentic

9.9%

TAU-2 Airline

agentic

74.1%

Briefcase v1.1

strategy

883.4 Elo

GDPval-AA v2.1

strategy

1186.3 Elo

AA-LCR v1.1

reasoning

79.7%

AA-Omniscience Index

reasoning

3.3

Architectural Profile & Capabilities

SWE-Bench Pro: 57.2%, Artificial Analysis Intelligence Index: 54, Claw-Eval Multimodal: 23.8 (matches Sonnet 4.6), Video-MME: 87.7 (rivals Gemini 3 Pro), HLE: 48.0%, autonomous workflows with **1,000+ tool calls** | Complex software engineering, long-horizon agentic tasks, real-time video/audio analysis, multimodal chart interpretation, multi-step automated workflows. **42% fewer tokens than Kimi K2.6 for equivalent performance** — highly token-efficient for cost-sensitive long workflows

Architecture Type: Mixture of Experts (MoE)

Sparse Mixture-of-Experts routing: activates only a subset of parameter experts per token, delivering frontier-level intelligence with exceptional efficiency.

Recommended Workloads & Primary Use Cases

Everyday reasoning

Drafting

Fast problem solving

Operational Caveats & Boundaries

•Extended context pricing applies above 256K tokens: $2/M in, $6/M out.

ARMES Zero Data Retention

Calls to MiMo V2.5 Pro are routed through strict Zero Data Retention inference channels. Prompts and outputs are never stored, indexed, or monitored by Xiaomi.

View Privacy Architecture

More Models from Xiaomi

Compare other engines available in this laboratory.

View Roster (4)

MiMo-V2.6-ProEco Tier

MiMo-V2.6-Pro foundation intelligence model.

128K ctx$0.43/M

MiMo-V2.6-FlashFree Tier

MiMo-V2.6-Flash foundation intelligence model.

128K ctx$0.14/M

MiMo V2.5Free Tier

Claw-Eval General: 62.3 — strong everyday performance with full multimodal coverage | Everyday tasks at high throughput, native multimodal Q&A (image/audio/video), cost-efficient see-hear-act workflow

1M ctx$0.40/M

Architectural Profile & Capabilities

SWE-Bench Pro: 57.2%, Artificial Analysis Intelligence Index: 54, Claw-Eval Multimodal: 23.8 (matches Sonnet 4.6), Video-MME: 87.7 (rivals Gemini 3 Pro), HLE: 48.0%, autonomous workflows with **1,000+ tool calls** | Complex software engineering, long-horizon agentic tasks, real-time video/audio analysis, multimodal chart interpretation, multi-step automated workflows. **42% fewer tokens than Kimi K2.6 for equivalent performance** — highly token-efficient for cost-sensitive long workflows

Architecture Type: Mixture of Experts (MoE)

Sparse Mixture-of-Experts routing: activates only a subset of parameter experts per token, delivering frontier-level intelligence with exceptional efficiency.