modelgrep
M

Microsoft: Phi 4

microsoft/phi-4

166th smartest of 178Cheaper than 90% of paidJSON
Use via OpenRouter ↗
Intelligence
10.4
166th of 178
Design Elo
Speed
72
109th fastest
Latency
224ms
first token
Input price
$0.065
30th cheapest
Context
16K
16K max out

How it compares

Smarter than7%
of all ranked models
Faster than63%
of all ranked models
Cheaper than90%
of all ranked models

Overview

[Microsoft Research](/microsoft) Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed. At 14 billion...

Benchmarks

independent · via OpenRouter
Artificial Analysis9th percentile
Intelligence Index
10.4
Coding Index
11.2
Agentic Index
0.0
GPQA Diamond
57%
Humanity's Last Exam
4%
SciCode
26%
Tau²-Bench (agentic)
0%

Providers & pricing (2)

ProviderIn $/MOut $/MUptime
NextBitint4$0.065$0.140100%
DeepInfrabf16$0.070$0.140100%

Specifications

Context window16K
Max output16K
Knowledge cutoffJun 2024
Input modalitiestext
Output modalitiestext
Prompt caching
Cache read price
ModeratedNo

Phi 4 FAQ

How much does Phi 4 cost?

Phi 4 costs $0.065 per million input tokens and $0.140 per million output tokens via OpenRouter, making it 30th cheapest of 298 paid models.

How smart is Phi 4?

Phi 4 scores 10.4 on the Artificial Analysis Intelligence Index, ranking 166th of 178 benchmarked models, with a GPQA Diamond score of 57%.

How fast is Phi 4?

Phi 4 generates around 72 tokens per second with 224ms time-to-first-token (p50), the 109th fastest tracked model.

What is Phi 4's context window?

Phi 4 supports a 16K-token context window and can output up to 16K tokens. It accepts text input.

Compare head-to-head