Noometry

Model comparison

GPT-4 vs Phi-4

Phi-4 is the stronger model overall, scoring 31.2 to 29.1 on the Noometry Index.

Last verified . 25 shared benchmarks.

GPT-4 OpenAI

29.1

Rank #316 Confirmed

Phi-4 Microsoft

31.2

Rank #279 Confirmed

Summary

  • They share 25 benchmarks with published results for both. GPT-4 scores higher in 4 categories and Phi-4 in 4 categories; 6 gaps are clear of the uncertainty.
  • The widest gap is in knowledge, where Phi-4 leads 32.6 to 18.4.
  • The biggest single-benchmark swing is MATH Level 5: 23% for GPT-4 and 64.9% for Phi-4.
  • Phi-4 is cheaper at $0.07 / $0.14 per million input/output tokens, against $30 / $60 for GPT-4.
  • Phi-4 accepts more context: 128K tokens versus 8K.
  • Phi-4 has downloadable open weights; the other is API-only.

Side by side

GPT-4Phi-4
ProviderOpenAIMicrosoft
Noometry Index29.131.2
Released2023-03-142024-12-11
WeightsProprietaryOpen
Context window8K128K
Max output8K4K
Input $ / M tokens$30$0.07
Output $ / M tokens$60$0.14
Results tracked3837

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Phi-4 leads

GPT-4: 31.6 (#283), Phi-4: 34.4 (#239)

BenchmarkGPT-4Phi-4
BigCodeBench Instruct46%45.5%
LMArena Coding12541231
BigCodeBench Complete57.2%55.4%
WeirdML12.4%—
LiveBench Coding—30.7%
HumanEval+79.3%—

Agentic & Tool Use Not comparable

GPT-4: —, Phi-4: 22.8 (#128)

BenchmarkGPT-4Phi-4
Berkeley Function Calling Leaderboard—28.8%
BALROG—11.6%
METR Time Horizons36.1%—

Reasoning Too close to call

GPT-4: 17.8 (#289), Phi-4: 17.7 (#291)

BenchmarkGPT-4Phi-4
Chess Puzzles4%1%
LMArena Hard Prompts12411220
Epoch Capabilities Index125.89130.42
LiveBench Reasoning—47.8%
Mystery Game Puzzles12%—
DTBench62.7%—
LiveBench Data Analysis—45.2%
LMCA17.1%—
BIG-Bench Hard75.1%—
ForecastBench57.8—
HellaSwag95.3%—
LiveBench—41.6%
WinoGrande87.5%—

Math Phi-4 leads

GPT-4: 10.8 (#309), Phi-4: 20.8 (#285)

BenchmarkGPT-4Phi-4
OTIS Mock AIME 2024-20251.1%13.8%
LMArena Math12691246
MATH Level 523%64.9%
LiveBench Math—42%
GSM8K92%—

Knowledge Phi-4 leads

GPT-4: 18.4 (#282), Phi-4: 32.6 (#209)

BenchmarkGPT-4Phi-4
GPQA Diamond35.7%56.1%
LMArena Expert12111203
MMLU86.4%84.8%
Confabulations—29.4%
Vectara Hallucination Rate—3.7%
TriviaQA84.8%—

Multilingual GPT-4 leads

GPT-4: 40.6 (#215), Phi-4: 37.2 (#237)

BenchmarkGPT-4Phi-4
LMArena Non-English12461197
LMArena Chinese12421212
LMArena French12831224
LMArena German12511222
LMArena Japanese12091158
LMArena Korean11841151
LMArena Russian12511209
LMArena Spanish12611234

Instruction Following GPT-4 leads

GPT-4: 65.3 (#222), Phi-4: 60.4 (#251)

BenchmarkGPT-4Phi-4
LMArena Instruction Following12411201
LiveBench Instruction Following—58.4%

Long Context Too close to call

GPT-4: 37.7 (#212), Phi-4: 36.9 (#226)

BenchmarkGPT-4Phi-4
LMArena Longer Query12441217

Writing & Preference Phi-4 leads

GPT-4: 34.9 (#268), Phi-4: 40.5 (#244)

BenchmarkGPT-4Phi-4
LMArena Text12631217
LMArena Creative Writing12441182
LMArena Multi-Turn12571206
Short-Story Creative Writing—62.6%
EQ-Bench Creative Writing752—
LiveBench Language—25.6%

Frequently asked questions

Is GPT-4 better than Phi-4?

Phi-4 is the stronger model overall, scoring 31.2 to 29.1 on the Noometry Index.

Which is cheaper, GPT-4 or Phi-4?

Phi-4 is cheaper. It lists at $0.07 per million input tokens and $0.14 per million output tokens; GPT-4 lists at $30 and $60.

Is GPT-4 or Phi-4 better for coding?

Phi-4 scores higher on coding benchmarks: 34.4 versus 31.6 in the Noometry coding category.

Which has the bigger context window?

Phi-4 does, with 128K tokens against 8K.

How many benchmarks do GPT-4 and Phi-4 share?

25 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Phi-4 has 37.

Related comparisons

Go deeper