NOUMENOpen app
NOUMEN

xiaomi · Open weights

MiMo V2.5 Pro Thinking

Added Jun 3, 2026Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

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

  • Reasoning
  • Tools
  • Structured output
Input
Text
Output
Text
Reasoning effort
none · high
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.44
Output
$0.87
Cache read
$0.0036
Cache write
$0.00

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
131.1K
Average output (7 days)
1.4K tokens
Longer than 84% of 465 models

Side by side

MiMo V2.5 Pro Thinking compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

MiMo V2.5 Pro ThinkingxiaomiThis modelMiMo V2.6 ProxiaomiRemove
Providerxiaomixiaomi
Discovery tierOpen weightsOpen weights
AddedJun 3, 2026Sep 21, 2026
Limits
Context window1.05M1.05M
Max output131.07K131.07K
Average output1.4K tokens1.2K tokens
Pricing, USD per 1M tokens
Input$0.44$0.44
Output$0.87$0.87
Cache read$0.0036$0.0036
Cache write$0.00$0.00
Performance
Throughput25.9 tok/s34.0 tok/sBest
Time to first token5.5 s4.1 sBest
Request success (24h)99.9%Best99.8%
Benchmark indexes
Intelligence Index26.046.3Best
Coding Index60.2—
Capabilities
ReasoningYesYes
ToolsYesYes
VisionNoYes
Structured outputYesYes
Audio inputNoYes
Video inputNoYes
Reasoning effortnone, highnone, high

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • MiMo V2.5 Pro Thinking
  • MiMo V2.6 Pro

Intelligence Index

  • MiMo V2.5 Pro Thinking26.0
  • MiMo V2.6 Pro46.3

MiMo V2.5 Pro Thinking: Better than 64% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • MiMo V2.5 Pro Thinking: 60.2
    • MiMo V2.6 Pro: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • MiMo V2.5 Pro Thinking: 13.7%
    • MiMo V2.6 Pro: 58.6%Top

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • MiMo V2.5 Pro Thinking: 2.0%
    • MiMo V2.6 Pro: 25.4%Top

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • MiMo V2.5 Pro Thinking: 883 Elo
    • MiMo V2.6 Pro: 1,512 EloTop

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • MiMo V2.5 Pro Thinking: 1,121 Elo
    • MiMo V2.6 Pro: 1,683 EloTop

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • MiMo V2.5 Pro Thinking: 4.0%
    • MiMo V2.6 Pro: 19.2%Top

  • AA-LCRv1.1

    Reasoning across long documents

    • MiMo V2.5 Pro Thinking: 79.7%
    • MiMo V2.6 Pro: 86.3%Top

  • MLCR-AA

    Long-context reasoning over medical records

    • MiMo V2.5 Pro Thinking: 5.6%
    • MiMo V2.6 Pro: 12.8%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • MiMo V2.5 Pro Thinking: 35.7%
    • MiMo V2.6 Pro: 49.4%Top

  • GPQA Diamond

    Graduate-level science questions

    • MiMo V2.5 Pro Thinking: 86.6%
    • MiMo V2.6 Pro: —

  • IFBench

    Following precise output instructions

    • MiMo V2.5 Pro Thinking: 79.9%
    • MiMo V2.6 Pro: —

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • MiMo V2.5 Pro Thinking: 0.0%
    • MiMo V2.6 Pro: 34.8%Top

  • SciCode

    Scientific computing in Python

    • MiMo V2.5 Pro Thinking: 50.6%
    • MiMo V2.6 Pro: 60.9%Top

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • MiMo V2.5 Pro Thinking: 43.2%
    • MiMo V2.6 Pro: —

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • MiMo V2.5 Pro Thinking: 94.2%
    • MiMo V2.6 Pro: —

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • MiMo V2.5 Pro Thinking: 79.7%
    • MiMo V2.6 Pro: 86.3%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
25.9 tok/s
Faster than 16% of 250 models
Time to first token
5.5 s
Quicker than 30% of 250 models
Prompt cache hits
58.9%
About $0.18 per 1M input tokens with caching
Request success (24h)
99.9%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 98.9%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "xiaomi/mimo-v2.5-pro:thinking",
    "reasoning_effort": "none",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="xiaomi/mimo-v2.5-pro:thinking",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

xiaomi

Other xiaomi models, starting with the same family.