Phi-4 API Pricing

Microsoft · released 2024-12-12

Pricing

$0.125
input / 1M tokens
0.013¢ / 1K
$0.50
output / 1M tokens
0.050¢ / 1K
$0.219
blended / 1M tokens
0.022¢ / 1K

What it actually costs

WorkloadTokensCost
Short chat reply500 in / 300 out0.0212¢
RAG query4,000 in / 500 out0.075¢
Long document summary50,000 in / 1,000 out0.675¢
Code review of a 2k-line file30,000 in / 2,000 out0.475¢
1,000 chat replies/day for 30 days30,000 × (500 in / 300 out)$6.38

Cost = (input tokens ÷ 1,000,000 × $0.125) + (output tokens ÷ 1,000,000 × $0.50).

Benchmarks

5.9
Intelligence
#445 of 533
Coding
18.0
Math
#194 of 249
36
Tokens/sec
1.00s
Time to first token

Similar intelligence, lower price

ModelIntelligenceBlended $/1M
Llama 3.1 Instruct 8B6.9$0.028
Gemma 4 E4B (Reasoning)8.9$0.04
Sarvam 30B (high)6.6$0.047
Nova Micro5.9$0.061
Granite 4.1 8B6.6$0.063

Leaderboard neighbours

ModelIntelligenceBlended $/1M
Granite 4.1 3B5.9
LFM2 24B A2B5.9
Nova Micro5.9$0.061
Phi-45.9$0.219
Gemini 1.0 Ultra5.8
Gemma 3n E4B Instruct Preview (May '25)5.8
Mistral Large (Feb '24)5.8$6.00

More from Microsoft

ModelIntelligenceBlended $/1M
Phi-4 Mini Instruct6.3
Phi-3 Mini Instruct 3.8B5.8
Phi-4 Multimodal Instruct5.8

Count tokens & estimate cost for Phi-4 See Phi-4 on the leaderboard

Source: Artificial Analysis · last synced 2026-09-22

FAQ

Who develops Phi-4?

Phi-4 is developed by Microsoft, released 2024-12-12.

How much does Phi-4 cost per million tokens?

Phi-4 costs $0.125 per 1M input tokens and $0.50 per 1M output tokens ($0.219 blended), per Artificial Analysis's official pricing data.

How does Phi-4 rank on benchmarks?

Phi-4 scores 5.9 on the Artificial Analysis Intelligence Index, ranking #445 of 533 models we track.

Is there a cheaper model with similar intelligence to Phi-4?

Yes — Llama 3.1 Instruct 8B scores similar intelligence (6.9) at $0.028/1M blended, versus $0.219/1M for Phi-4.

Where does this data come from?

Pricing and benchmark data for Phi-4 comes from Artificial Analysis (artificialanalysis.ai), synced daily. Last synced 2026-09-22.