MultimodalOpen weights

Phi-4-multimodal-instruct

Microsoft

Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error

Specifications

Type: Multimodal
Access: Open weights
Released: February 1, 2025
License: MIT
Parameters: 5600000000
Context window: 128,000 tokens
Knowledge cutoff: June 1, 2024
Output speed: 25 tok/s
Latency (TTFT): 0.5s
API pricing: $0.05 in · $0.10 out / 1M tokens

Capabilities

Function callingStructured output

Benchmarks

Reasoning

31.5

Coding

13.1

Math

69.3

Multimodal

AI2D	82.3
ChartQA	81.4
DocVQA	93.2
MathVista	62.4
MMMU	55.1
MMMU-Pro	38.5

General

MMLU-Pro	48.5
Humanity’s Last Exam	4.4

Compare Phi-4-multimodal-instruct with

vs Sonar Reasoning Pro vs R1 1776 vs Qwen3.7 Max vs Gemini 3.5 Flash

See all Phi-4-multimodal-instruct alternatives →