Skip to content
VECTOR WIREAI INTELLIGENCE
⌘K
UTC
Models741Deals30d271Policy30d66, 14 pendingSources1,015

Phi 4

MicrosoftReleased Dec 12, 2024Current

Score basis

Phi 4 is behind the leaders in factuality. Too few results yet to rate reasoning, coding, agentic tasks, safety, long context, math, multimodal tasks, multilingual tasks, or instruction following.

Capability profile

Bars show the model's median result as a share of the leading model's, per capability. Select a row to see its results.

Limited1 capability
  1. Factuality−35.9%3 of 4
Not rated9 capabilities

Too few results yet to rate Reasoning, Coding, Agentic, Safety, Long Context, Math, Multimodal, Multilingual or Instruction Following.

Price

$0.13input$0.50outputper million tokens

From Artificial Analysis · 3 providers tracked · All prices

Cheaper than 80% of 330 priced models · 3:1 input-to-output blend, log scale

Evidence

36results on33benchmarks

  • 10 independently verified
  • 14 aggregator
  • 12 vendor-reported

From 8 sources · latest Oct 8, 2026 · How verification works

API features

Structured outputs

As listed by OpenRouter

Research

30 papers reference Phi 4

Phi 4 benchmark results

36 results on 33 benchmarks, grouped by capability. Each bar is the result as a share of the capability leader's; every result links to its source.

Factuality

LimitedFull factuality ranking

35.9% behind the leader3 of 4 ranked benchmarks measured

Reasoning

Not enough dataFull reasoning ranking

0 of 6 ranked benchmarks measured

Show 1 more reasoning resultHide 1 reasoning result

Coding

Not enough dataFull coding ranking

0 of 10 ranked benchmarks measured

Long Context

Not enough dataFull long context ranking

0 of 3 ranked benchmarks measured

Instruction Following

Not enough dataFull instruction following ranking

0 of 3 ranked benchmarks measured

More results

Benchmarks outside the capability baskets. They are not ranked against a leader.

Show 17 more resultsHide 17 results

Phi 4: common questions

Who makes Phi 4?

Phi 4 is made by Microsoft.

When was Phi 4 released?

Phi 4 was released on Dec 12, 2024, according to Artificial Analysis.

What is Phi 4 good at?

Phi 4 is behind the leaders in factuality. Too few results yet to rate reasoning, coding, agentic tasks, safety, long context, math, multimodal tasks, multilingual tasks, or instruction following.

How much does Phi 4 cost?

Phi 4 costs $0.13 per million input tokens and $0.50 per million output tokens, according to Artificial Analysis. We track its price at 3 providers. At a mix of three input tokens to one output token, it is cheaper than 80% of the 330 priced models we track.

How many benchmarks has Phi 4 been tested on?

We track 36 results for Phi 4 on 33 benchmarks from 8 sources, 10 of them independently verified. The latest was recorded on Oct 8, 2026.

Which API features does Phi 4 support?

OpenRouter lists structured outputs for Phi 4.

About this record

Where Phi 4's numbers come from, and every name it appears under.

Tracked since
Apr 27, 2026
Newest source mention
May 26, 2026

Where the results come from

Verification: 36 scores · 10 independently verified · 14 aggregator-attributed · 12 vendor-reported. How these tiers are assigned

From 8 sources on 5 sites. Artificial Analysis supplies 16 of them; the 10 independently verified results come from 4 sites. Bars are coloured by trust tier.

  • artificialanalysis.ai16
  • api.llm-stats.com10
  • huggingface.co5
  • raw.githubusercontent.com4
  • arxiv.org1

Also known as

phi-4 (Reasoning)phi-4 (Non-Reasoning)microsoft/phi-4