LLMversus
Microsoft

Phi-4

Complete specs, pricing, and benchmark data for Phi-4 by Microsoft. Last verified 2026-04-20.

JSON ModeStreaming
Pricing

Input / 1M tokens

$0.065

Output / 1M tokens

$0.140

Cached Input / 1M

$0.018

Context & Output

Context Window

16.384K

Max Output

4,096

TTFT

100ms

Speed

160 tok/s

Benchmarks

Arena ELO

1150

Coding ELO

1130

Reasoning ELO

1140

HumanEval

80

MMLU

80.5

MATH

72

GPQA

45

Try Phi-4 API

Start building with Phi-4

Get API Access→

Price History (Input $/M tokens)

Strengths
  • +Ultra-low cost for a capable model
  • +Strong math for its size (14B params)
  • +Very fast inference
  • +Can run on consumer hardware
Limitations
  • -Small 16K context window
  • -No vision or function calling
  • -Limited compared to larger models on complex tasks

Best For

Cost SensitiveEdge DeploymentMathLightweight Tasks

Compare Phi-4 with...

vs Claude Opus 4vs Gemini 2.5 Provs o3vs DeepSeek R1vs o1vs Qwen 3 235B MoEvs Gemini Experimental 1206vs GPT-4.5vs DeepSeek R1 (Groq)vs Llama 4 Maverickvs DeepSeek R1 (Together)vs Grok 3
Official Pricing Page →
Your ad here