Browse all Xiaomi text models
Provider logo

MiMo V2.5 Thinking

xiaomi/mimo-v2.5:thinking
Provider logo

MiMo V2.5 Thinking

xiaomi/mimo-v2.5:thinking

MiMo V2.5 with Xiaomi thinking enabled. It supports deep reasoning, tool calling, structured outputs, and web search with up to 1M context.

Added Jun 3, 2026

Model weights

Context Window

1.0M

Max Output

131.1K

Avg output tokens (7d)

747 tokens

61%

Input Price (Auto)

$0.12/1M

Output Price (Auto)

$0.24/1M

Cache Read (Auto)

$0.0026/1M

Capabilities

Benchmarks

Performance metrics and benchmarks

Sourced from Artificial Analysis.

Intelligence Index

22.3

Better than 72% of models compared

Coding Index

56.8

Better than 68% of models compared

Agentic Index

17.4

Better than 52% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

18.4%

Better than 45% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

751 Elo

Better than 39% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

1079 Elo

Better than 54% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

4.0%

Better than 28% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

73.0%

Better than 72% of models compared

Reasoning

HLE

Humanity's Last Exam

27.2%

Better than 78% of models compared

IFBench

Instruction-following benchmark

67.1%

Better than 80% of models compared

CritPt

Research-level physics reasoning

3.7%

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

0.0%

Better than 17% of models compared

SciCode

Python programming for scientific computing

43.9%

Better than 39% of models compared

Knowledge

AA-Omniscience Accuracy

Proportion of correctly answered questions

16.8%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

31.9%

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

84.9%

Better than 78% of models compared

Terminal-Bench Hard (legacy)

Agentic coding and terminal use

41.7%

Better than 88% of models compared

T²-Bench Telecom (legacy)

Conversational AI agents in dual-control scenarios

90.6%

Better than 85% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

73.0%

Better than 72% of models compared

GDPval-AA (unversioned / legacy)

Economically valuable tasks

29.0%

Last updated Sep 12, 2026

Artificial Analysis

Providers

Choose explicit providers for this model. Auto routing remains available as the default option.

Loading provider options…

Compare MiMo V2.5 Thinking with similar models from the same provider or model family.

MiMo V2.5 Pro (Crof)

xiaomi/mimo-v2.5-pro-crof

MiMo V2.5 Pro is Xiaomi's long-context flagship general model for coding and agentic orchestration. This separately served variant is intended for users concerned about censorship on the regular Xiaomi MiMo V2.5 Pro.

MiMo V2.5 Pro Thinking (Crof)

xiaomi/mimo-v2.5-pro-crof:thinking

MiMo V2.5 Pro with Xiaomi thinking enabled for coding, long-context reasoning, and agentic orchestration. This separately served thinking variant is intended for users concerned about censorship on the regular Xiaomi MiMo V2.5 Pro.

MiMo V2.5 Pro Thinking

xiaomi/mimo-v2.5-pro:thinking

MiMo V2.5 Pro with Xiaomi thinking enabled for coding, long-context reasoning, and agentic orchestration.

MiMo V2.5

xiaomi/mimo-v2.5

MiMo V2.5 is Xiaomi's full-modal understanding model for agent workflows. It supports tool calling, structured outputs, and web search with up to 1M context.

MiMo V2.5 Pro

xiaomi/mimo-v2.5-pro

MiMo V2.5 Pro is Xiaomi's long-context flagship general model for coding and agentic orchestration. It supports tool calling and structured outputs with up to 1M context.

Schematron V2 Small

inference-net/schematron-v2-small

Inference.net's 3B-parameter HTML-to-JSON extraction model, focused on accuracy for complex schemas and long web pages. It turns HTML into typed, structured data for web scraping and product catalog ingestion, with a 128K-token context window. Supply HTML in the user message and extraction instructions in a JSON schema via response_format; it does not follow ordinary chat or system prompts.