NOUMENOpen app
NOUMEN

microsoft · Open weights

Phi 4 Mini

Added Jul 26, 2025Open weights

Compare models

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

Loading models…

Use in the app

Phi 4 Mini by Microsoft. A small multilingual model.

Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.17
Output
$0.68

Reference prices as published by the model provider. API example

Specifications

Context window
128K
Max output
16.4K
Average output (7 days)
24 tokens
Longer than 0% of 468 models

Side by side

Phi 4 Mini 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…

Phi 4 MinimicrosoftThis modelCeleris 1celerisRemove
Providermicrosoftceleris
Discovery tierOpen weightsBalanced
AddedJul 26, 2025Jul 25, 2026
Limits
Context window128KBest8.19K
Max output16.38KBest8.19K
Average output24 tokens—
Pricing, USD per 1M tokens
Input$0.17Best$2.00
Output$0.68Best$6.00
Performance
Request success (24h)100.0%—
Benchmark indexes
Intelligence Index6.36.3
Coding Index3.814.4Best
Math Index6.7—
Capabilities
Reasoning effort——

Public evaluations

Benchmarks

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

  • Phi 4 Mini
  • Celeris 1

Intelligence Index

  • Phi 4 Mini6.3
  • Celeris 16.3

Phi 4 Mini: Better than 4% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Phi 4 Mini: 3.8
    • Celeris 1: 14.4Top

  • Math Index

    Composite of the math evaluations

    • Phi 4 Mini: 6.7
    • Celeris 1: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Phi 4 Mini: —
    • Celeris 1: 0.4%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Phi 4 Mini: —
    • Celeris 1: 0.2%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Phi 4 Mini: —
    • Celeris 1: 151 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Phi 4 Mini: —
    • Celeris 1: 301 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Phi 4 Mini: —
    • Celeris 1: 1.2%

  • AA-LCRv1.1

    Reasoning across long documents

    • Phi 4 Mini: 15.3%
    • Celeris 1: 38.0%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Phi 4 Mini: 4.4%
    • Celeris 1: 6.8%Top

  • GPQA Diamond

    Graduate-level science questions

    • Phi 4 Mini: 33.1%
    • Celeris 1: 63.1%Top

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Phi 4 Mini: 46.5%
    • Celeris 1: 78.0%Top

  • IFBench

    Following precise output instructions

    • Phi 4 Mini: 21.1%
    • Celeris 1: —

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Phi 4 Mini: —
    • Celeris 1: 0.0%

  • SciCode

    Scientific computing in Python

    • Phi 4 Mini: —
    • Celeris 1: 21.6%

  • LiveCodeBench

    Recent competitive programming problems

    • Phi 4 Mini: 12.6%
    • Celeris 1: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Phi 4 Mini: 6.7%
    • Celeris 1: —

  • AIME

    Competition mathematics

    • Phi 4 Mini: 3.0%
    • Celeris 1: —

  • MATH-500

    Competition math problems

    • Phi 4 Mini: 69.6%
    • Celeris 1: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Phi 4 Mini: 0.0%
    • Celeris 1: —

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Phi 4 Mini: 8.2%
    • Celeris 1: —

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Phi 4 Mini: 15.3%
    • Celeris 1: 38.0%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.

Prompt cache hits
0.0%
Request success (24h)
100.0%
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 100.0%
  • 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": "phi-4-mini-instruct",
    "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="phi-4-mini-instruct",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

microsoft

Other microsoft models, starting with the same family.