Noometry

Model comparison

gpt-oss-20b vs Phi-4

gpt-oss-20b is the stronger model overall, scoring 32.5 to 31.2 on the Noometry Index.

Last verified . 20 shared benchmarks.

gpt-oss-20b OpenAI

32.5

Rank #255 Confirmed

Phi-4 Microsoft

31.2

Rank #279 Confirmed

Summary

  • They share 20 benchmarks with published results for both. gpt-oss-20b scores higher in 7 categories and Phi-4 in 2 categories; 9 gaps are clear of the uncertainty.
  • The widest gap is in math, where gpt-oss-20b leads 39.4 to 20.8.
  • The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 65.3% for gpt-oss-20b and 13.8% for Phi-4.
  • gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.07 / $0.14 for Phi-4.
  • gpt-oss-20b accepts more context: 131K tokens versus 128K.

Side by side

gpt-oss-20bPhi-4
ProviderOpenAIMicrosoft
Noometry Index32.531.2
Released2025-08-052024-12-11
WeightsOpenOpen
Context window131K128K
Max output16K4K
Input $ / M tokens$0.018$0.07
Output $ / M tokens$0.09$0.14
Results tracked3437

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

Category by category

Coding gpt-oss-20b leads

gpt-oss-20b: 37.6 (#192), Phi-4: 34.4 (#239)

Benchmarkgpt-oss-20bPhi-4
LMArena Coding13061231
SciCode34.4%—
WeirdML40.9%—
BigCodeBench Instruct—45.5%
LiveBench Coding—30.7%
BigCodeBench Complete—55.4%
ALE-Bench566.05—

Agentic & Tool Use Phi-4 leads

gpt-oss-20b: 9.3 (#154), Phi-4: 22.8 (#128)

Benchmarkgpt-oss-20bPhi-4
Terminal-Bench3.4%—
Berkeley Function Calling Leaderboard—28.8%
BALROG—11.6%

Reasoning gpt-oss-20b leads

gpt-oss-20b: 19.3 (#261), Phi-4: 17.7 (#291)

Benchmarkgpt-oss-20bPhi-4
Chess Puzzles4%1%
LMArena Hard Prompts12741220
Epoch Capabilities Index137.82130.42
Kagi LLM Benchmark53.2%—
CritPt1.4%—
LiveBench Reasoning—47.8%
DTBench68%—
LiveBench Data Analysis—45.2%
LMCA14.5%—
LiveBench—41.6%

Math gpt-oss-20b leads

gpt-oss-20b: 39.4 (#103), Phi-4: 20.8 (#285)

Benchmarkgpt-oss-20bPhi-4
OTIS Mock AIME 2024-202565.3%13.8%
LMArena Math13171246
Omni-MATH56.5%—
LiveBench Math—42%
MATH Level 5—64.9%

Knowledge gpt-oss-20b leads

gpt-oss-20b: 34.6 (#195), Phi-4: 32.6 (#209)

Benchmarkgpt-oss-20bPhi-4
GPQA Diamond60.8%56.1%
LMArena Expert12581203
MMLU-Pro74%—
Confabulations—29.4%
Vectara Hallucination Rate—3.7%
GPQA (HELM)59.4%—
MMLU—84.8%

Multilingual gpt-oss-20b leads

gpt-oss-20b: 42.2 (#197), Phi-4: 37.2 (#237)

Benchmarkgpt-oss-20bPhi-4
LMArena Non-English12681197
LMArena Chinese13141212
LMArena German12551222
LMArena Japanese12441158
LMArena Korean12361151
LMArena Russian12781209
LMArena Spanish12671234
LMArena French—1224

Instruction Following gpt-oss-20b leads

gpt-oss-20b: 61.8 (#240), Phi-4: 60.4 (#251)

Benchmarkgpt-oss-20bPhi-4
LMArena Instruction Following12361201
LiveBench Instruction Following—58.4%
IFEval73.2%—

Long Context gpt-oss-20b leads

gpt-oss-20b: 37.9 (#209), Phi-4: 36.9 (#226)

Benchmarkgpt-oss-20bPhi-4
LMArena Longer Query12501217

Writing & Preference Phi-4 leads

gpt-oss-20b: 35.5 (#265), Phi-4: 40.5 (#244)

Benchmarkgpt-oss-20bPhi-4
LMArena Text12871217
LMArena Creative Writing12011182
LMArena Multi-Turn12681206
Short-Story Creative Writing—62.6%
EQ-Bench Creative Writing666—
WildBench73.7%—
LiveBench Language—25.6%

Frequently asked questions

Is gpt-oss-20b better than Phi-4?

gpt-oss-20b is the stronger model overall, scoring 32.5 to 31.2 on the Noometry Index.

Which is cheaper, gpt-oss-20b or Phi-4?

gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; Phi-4 lists at $0.07 and $0.14.

Is gpt-oss-20b or Phi-4 better for coding?

gpt-oss-20b scores higher on coding benchmarks: 37.6 versus 34.4 in the Noometry coding category.

Which has the bigger context window?

gpt-oss-20b does, with 131K tokens against 128K.

How many benchmarks do gpt-oss-20b and Phi-4 share?

20 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and Phi-4 has 37.

Related comparisons

Go deeper