Phi 4 Multimodal Instruct

Phi-4-multimodal-instruct 是一个轻量级（55.7亿参数）的开放多模态基础模型，它利用了 Phi-3.5 和 4.0 的研究成果和数据集。它可以处理文本、图像和音频输入，并生成文本输出，支持 128K 个标记的上下文长度。通过 SFT、DPO 和 RLHF 进行增强，以提高指令遵循能力和安全性。

Intelligence(Relatively Weak)

Speed(Slow)

Input Supported Modalities

Yes

Is Reasoning Model

128,000

Context Window

128,000

Maximum Output Tokens

2024-06-01

Knowledge Cutoff

Go Compare

Pricing

￥0.36 /M tokens

Input

￥0.72 /M tokens

Output

- /M tokens

Blended Price

Quick Simple Comparison

Input

Output

Phi-3 Medium Instruct 14B

￥0.1

Phi-4 Mini Instruct

Phi-4 Multimodal Instruct

￥0.05

Basic Parameters

Phi-4 Multimodal InstructTechnical Parameters

Parameter Count

5,600.0M

Context Length

128.00k tokens

Training Data Cutoff

2024-06-01

Open Source Category

Open Weights (Permissive License)

Multimodal Support

Text, Image

Throughput

Release Date

2025-02-26

Response Speed

21.646,065 tokens/s

Benchmark Scores

Below is the performance of Phi-4 Multimodal Instruct in various standard benchmark tests. These tests evaluate the model's capabilities in different tasks and domains.

Intelligence Index

Large Language Model Intelligence Level

26.91

Large Language Model Intelligence Level

Coding Index

Indicator of AI model performance on coding tasks

12.06

Indicator of AI model performance on coding tasks

Math Index

Capability indicator in solving mathematical problems, mathematical reasoning, or performing math-related tasks

39.3

Capability indicator in solving mathematical problems, mathematical reasoning, or performing math-related tasks

MMLU Pro

Massive Multitask Multimodal Understanding - Testing understanding of text, images, audio, and video

48.5

Massive Multitask Multimodal Understanding - Testing understanding of text, images, audio, and video

GPQA

Graduate Physics Questions Assessment - Testing advanced physics knowledge with diamond science-level questions

31.5

Graduate Physics Questions Assessment - Testing advanced physics knowledge with diamond science-level questions

HLE

The model's comprehensive average score on the Hugging Face Open LLM Leaderboard

4.4

The model's comprehensive average score on the Hugging Face Open LLM Leaderboard

LiveCodeBench

Specific evaluation focused on assessing large language models' ability in real-world code writing and solving programming competition problems

13.1

Specific evaluation focused on assessing large language models' ability in real-world code writing and solving programming competition problems

SciCode

The model's capability in code generation for scientific computing or specific scientific domains

HumanEval

Score achieved by the AI model on the specific HumanEval benchmark test set

72.9

Score achieved by the AI model on the specific HumanEval benchmark test set

Math 500 Score

Score on the first 500 larger, more well-known mathematical benchmark tests

69.3

Score on the first 500 larger, more well-known mathematical benchmark tests

AIME Score

An indicator measuring an AI model's ability to solve high-difficulty mathematical competition problems (specifically AIME level)

9.3

An indicator measuring an AI model's ability to solve high-difficulty mathematical competition problems (specifically AIME level)

Output tokens/million

Output tokens/million

Output tokens/million

Output tokens/million

Output tokens/million

131k

Context Length

Gemini 2.0 Flash Lite (Preview)

Output tokens/million

Output tokens/million

33k

Context Length

Qwen2.5 Coder Instruct 32B

Output tokens/million

131k

Context Length

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご