AIbase
EN
Home
AI Models Finder
Phi 3 Medium Instruct 14B
P

Phi 3 Medium Instruct 14B

Phi-3.5-mini-instruct is a 3.8B parameter model that supports up to 128K context tokens and has improved multilingual capabilities across more than 20 languages. It has received additional training and safety post-training to enhance instruction following, reasoning, mathematics, and code generation. It uses the MIT license and is well-suited for environments with memory or latency constraints.
Intelligence(Relatively Weak)
Speed(Relatively Slow)
Input Supported Modalities
No
Is Reasoning Model
128,000
Context Window
128,000
Maximum Output Tokens
2023-10-01
Knowledge Cutoff
Go Compare

Pricing

￥0.72 /M tokens
Input
￥0.72 /M tokens
Output
￥2.14 /M tokens
Blended Price

Quick Simple Comparison

Input

Output

Phi-4 Mini Instruct
Phi-4 Multimodal Instruct
￥0.05
Phi-3 Medium Instruct 14B
￥0.1

Basic Parameters

Phi-3 Medium Instruct 14BTechnical Parameters
Parameter Count
3,800.0M
Context Length
128.00k tokens
Training Data Cutoff
2023-10-01
Open Source Category
Open Weights (Permissive License)
Multimodal Support
Text Only
Throughput
23
Release Date
2024-05-21
Response Speed
52.418,564 tokens/s

Benchmark Scores

Below is the performance of Phi-3 Medium Instruct 14B in various standard benchmark tests. These tests evaluate the model's capabilities in different tasks and domains.
Intelligence Index
24.54
Large Language Model Intelligence Level
Coding Index
13.42
Indicator of AI model performance on coding tasks
Math Index
23.8
Capability indicator in solving mathematical problems, mathematical reasoning, or performing math-related tasks
MMLU Pro
54.3
Massive Multitask Multimodal Understanding - Testing understanding of text, images, audio, and video
GPQA
32.6
Graduate Physics Questions Assessment - Testing advanced physics knowledge with diamond science-level questions
HLE
4.5
The model's comprehensive average score on the Hugging Face Open LLM Leaderboard
LiveCodeBench
15
Specific evaluation focused on assessing large language models' ability in real-world code writing and solving programming competition problems
SciCode
11.8
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
Math 500 Score
46.3
Score on the first 500 larger, more well-known mathematical benchmark tests
AIME Score
1.3
An indicator measuring an AI model's ability to solve high-difficulty mathematical competition problems (specifically AIME level)
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase