P

Phi 3 Mini 4k Instruct Onnx Web

Developed by microsoft
Phi-3 Mini-4K-Instruct ONNX model for in-browser inference, accelerated via ONNX Runtime Web for browser-based processing.
Downloads 243
Release Time : 5/17/2024

Model Overview

A lightweight, state-of-the-art open-source model with 3.8 billion parameters, focusing on high-quality and inference-intensive features, suitable for tasks like commonsense reasoning, language understanding, mathematics, coding, long-context, and logical reasoning.

Model Features

In-browser inference
Runs entirely in the browser without server support, accelerated via ONNX Runtime Web.
Lightweight and high-performance
A lightweight model with 3.8 billion parameters, demonstrating strong and state-of-the-art performance among models with fewer than 13 billion parameters.
WebGPU acceleration
Recommended to use the WebGPU backend for efficient operation, achieving approximately 42 tokens per second on an NVIDIA GeForce RTX 4090.
Optimized storage
The model is in fp16 format with int4 block-quantized weights, ensuring the model and external data files remain under 2GB for easy caching in Chromium.

Model Capabilities

Text generation
Commonsense reasoning
Language understanding
Mathematical computation
Code generation
Logical reasoning

Use Cases

Conversational systems
Smart chatbot
Build conversational AI applications that run entirely in the browser.
Low-latency interactive experience
Education
Learning assistant
Help students solve problems in mathematics, programming, etc.
Provide instant and accurate answers
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase