ZMIME
Microsoft · Released Feb 26, 2025

Phi-4-multimodal-instruct

Compact Microsoft multimodal model for text, image, and speech understanding.

  • Open weights
  • Tool calling
  • Vision
  • Audio input
Capability
—No ECI score yet
Input
$0.07per 1M tokens
Output
$0.11per 1M tokens
Context
131K131K max output
Pricing

What it costs

Input
$0.07per million tokens
Output
$0.11per million tokens
Blended (3:1)
$0.08cheaper than 92% of priced models

Typical monthly bills

Monthly usageEstimated cost
Side project2M input + 0.5M output tokens$0.195
Team assistant25M input + 5M output tokens$2.30
Production app250M input + 50M output tokens$23.00

Median of 1 paid API provider on models.dev.

Capability

Independent benchmarks

Epoch AI has not published a Capabilities Index score for Phi-4-multimodal-instruct yet.

Specifications

The details

Lab
Microsoft
Released
Feb 26, 2025
Knowledge cutoff
Jun 2024
Context window
131,072 tokens
Max output
131,072 tokens
Inputs
Text, Images, Audio
Output
Text
Reasoning
No
Tool calling
Yes
Structured output
No
Availability
1 API providerlisted on models.dev
Same lab

More from Microsoft

Questions

About Phi-4-multimodal-instruct

How much does Phi-4-multimodal-instruct cost?

Phi-4-multimodal-instruct costs $0.07 input / $0.11 output per million tokens (median across 1 API provider). At a 3:1 input-to-output mix that is $0.08 per million tokens, cheaper than 92% of the 445 priced models we track.

What is the context window of Phi-4-multimodal-instruct?

Phi-4-multimodal-instruct accepts up to 131,072 tokens per request and can write up to 131,072 tokens in one response.

How good is Phi-4-multimodal-instruct?

Epoch AI has not published a Capabilities Index score for Phi-4-multimodal-instruct yet.

Is Phi-4-multimodal-instruct open source?

Yes. Microsoft publishes the weights under the MIT licence, so it can be downloaded and self-hosted.

What inputs does Phi-4-multimodal-instruct support?

Phi-4-multimodal-instruct accepts text, images and audio and replies in text. It is not a dedicated reasoning model, supports tool calling.

When was Phi-4-multimodal-instruct released?

Microsoft released Phi-4-multimodal-instruct on Feb 26, 2025. Its training data runs to Jun 2024.

What are the best alternatives to Phi-4-multimodal-instruct?

The closest current models from other labs on capability, price and release date are Command R7B Arabic (Cohere, $0.037 / $0.15), Reka Flash 3 (Reka AI, $0.10 / $0.20), Qwen-VL OCR (Alibaba (Qwen), $0.07 / $0.16) and Nova Micro (Amazon, $0.035 / $0.14).