Phi-3 Vision: Complete Benchmark Performance Guide

other phi-3-vision 2024-05-21 MIT

In-depth analysis of Phi-3 Vision performance on MMLU, HumanEval, GSM8K, MATH, BBH benchmarks, including pricing and use cases.

Key Specifications

Vendorother
Versionphi-3-vision
Release Date2024-05-21
Context Window4096 tokens
Input Modalitiestext, image
Output Modalitiestext
LicenseMIT
Documentationhttps://huggingface.co/models

Benchmark Performance

Benchmark Score Unit Evaluated At Notes Source
MMLU 56.4 % 2024-05-21 5-shot view
HUMANEVAL 48.2 pass@1 2024-05-21 — view
GSM8K 30 % 2024-05-21 0-shot CoT view
MATH 24.3 % 2024-05-21 0-shot CoT view
BBH 40.1 % 2024-05-21 3-shot CoT view
GPQA 17.4 % 2024-05-21 0-shot view
IFEVAL 56.4 % 2024-05-21 prompt_strict view
ARC 77.4 % 2024-05-21 challenge view
MUSR 32.1 % 2024-05-21 0-shot view
WINOGRANDE 74.7 % 2024-05-21 0-shot view

Pricing

Tier Price Currency
Input$0.2 / MtokUSD
Output$0.2 / MtokUSD
Cache Read$0 / MtokUSD
Cache Write$0 / MtokUSD

Source: https://huggingface.co/models · as of 2024-05-21

Compliance

  • Data Residency: self-host
  • SOC2: ✗
  • HIPAA: ✗
  • GDPR: ✗
  • ISO 27001: ✗

Phi-3 Vision

Model Overview

Microsoft Phi-3 Vision 4.2B 多模态模型, 4K 上下文, 支持图像理解, 在 4B 规模上多模态能力领先。

Core Specifications

Vendor Version Release Date Context Window Input Modalities Output Modalities License
Other phi-3-vision 2024-05-21 4K text, image text MIT

Benchmark Performance

Benchmark Score Unit Notes
MMLU (Massive Multitask Language Understanding) 56.4 % 5-shot
HumanEval 48.2 pass@1 —
GSM8K (Grade School Math 8K) 30.0 % 0-shot CoT
MATH 24.3 % 0-shot CoT
BBH (BIG-Bench Hard) 40.1 % 3-shot CoT
GPQA 17.4 % 0-shot
IFEval 56.4 % prompt_strict
ARC 77.4 % challenge
MUSR 32.1 % 0-shot
WinoGrande 74.7 % 0-shot

Pricing

Input Output Cache Read Cache Write
— — — —

per million tokens

Strengths

  • Input Modalities: text, image, audio.

Weaknesses

  • MMLU 56.4, weak knowledge reasoning.
  • HumanEval 48.2, coding weak.
  • Proprietary, not self-hostable.
  • Context window 4K is limited.

Use Cases

  • Vision and image understanding

References