Enter opens the highlighted result. Esc closes.

GPT-5 vs Phi 4

Updated October 21, 2026. 44 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace GPT-5 Replace Phi 4
GPT-5OpenAI
Coding
38.8
Overall
23.0
Output $/M
$10.00
44
Phi 4Microsoft
Coding
No data
Overall
5.9
Output $/M
$0.14
0
vs boards won of 44 shared

Decision reading

Only GPT-5 has a coding score, 38.8 on the Vibe Coding Index; Phi 4 has none, so the two cannot be ranked on coding.

Phi 4 has the edge where output tokens cost 71.4x less.

Model AGPT-5OpenAI38.8/1003 independent benchmarks$0.39 per solved taskModel BPhi 4MicrosoftNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
44
GPT-5 only
65
Phi 4 only
1
Like-for-like categories
7 of 9

Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
GPT-5 38.8 from 3 benchmarks, Phi 4 not benchmarked.
Widest category
Agentic work, 53.9 points apart on average over 2 shared benchmarks.
Widest single gap
τ²-Bench Telecom: GPT-5 86.5% against Phi 4 0.0%, 86.5 points apart on a 0 to 100 scale.
Output price
GPT-5 $10.00 against Phi 4 $0.14 per million output tokens, 71.4x apart.
Context window
GPT-5 takes 400K tokens, Phi 4 16K.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryGPT-5BarsPhi 4Ties
Coding40
Intelligence20
Agentic work20
Reasoning60
Math60
Knowledge50
General190
All shared boards440
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
Coding benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
Aider polyglot88.0%Not measuredOnly GPT-5 measured
SWE-bench Verified74.4%Not measuredOnly GPT-5 measured
LMArena WebDev1418Not measuredOnly GPT-5 measured
AA Coding Index37.8Not measuredOnly GPT-5 measured
LiveCodeBench (Vals AI)85.9%Not measuredOnly GPT-5 measured
SWE-bench Verified (Vals AI)69.0%Not measuredOnly GPT-5 measured
Vibe Code Bench (Vals AI)20.1%Not measuredOnly GPT-5 measured
LiveCodeBench (Artificial Analysis run)84.6%23.1%GPT-5
Terminal-Bench Hard (Artificial Analysis run)37.9%3.8%GPT-5
Terminal-Bench49.6%Not measuredOnly GPT-5 measured
GSO-Bench6.9%Not measuredOnly GPT-5 measured
WeirdML60.7%Not measuredOnly GPT-5 measured
ALE-Bench807.6Not measuredOnly GPT-5 measured
LMArena Text: coding14361231GPT-5
LMArena Text: software and IT14201228GPT-5
LMArena WebDev: HTML1427Not measuredOnly GPT-5 measured
Neither model has a result on the other 22 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
AA Intelligence Index23.05.9GPT-5
Epoch Capabilities Index150.0130.5GPT-5
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
Agentic work benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
τ²-Bench Telecom86.5%0.0%GPT-5
METR time horizons69.6%Not measuredOnly GPT-5 measured
Remote Labor Index1.7%Not measuredOnly GPT-5 measured
GDPval34.8%Not measuredOnly GPT-5 measured
BALROG32.8%11.6%GPT-5
DeepResearch Bench49.6%Not measuredOnly GPT-5 measured
Neither model has a result on the other 33 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index60
Reasoning benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
GPQA Diamond85.6%Not measuredOnly GPT-5 measured
MMMU Pro (Vals AI)81.5%Not measuredOnly GPT-5 measured
GPQA Diamond (Artificial Analysis run)85.4%57.5%GPT-5
Humanity's Last Exam (Artificial Analysis run)28.5%3.8%GPT-5
GPQA Diamond (Epoch AI run)86.2%56.1%GPT-5
ARC-AGI65.7%Not measuredOnly GPT-5 measured
ARC-AGI-29.9%Not measuredOnly GPT-5 measured
SimpleBench56.7%Not measuredOnly GPT-5 measured
Chess puzzles36.0%Not measuredOnly GPT-5 measured
Mystery game puzzles23.0%Not measuredOnly GPT-5 measured
EBR-bench12.7%Not measuredOnly GPT-5 measured
VPCT66.0%Not measuredOnly GPT-5 measured
LMArena Text: hard prompts14151220GPT-5
LMArena Text: expert14191203GPT-5
LMArena Text: hard prompts in English14191230GPT-5
LMArena Vision: diagrams1252Not measuredOnly GPT-5 measured
LMArena Vision: homework1278Not measuredOnly GPT-5 measured
Neither model has a result on the other 4 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index60
Math benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
AIME93.4%Not measuredOnly GPT-5 measured
MATH 500 (Vals AI)96.0%Not measuredOnly GPT-5 measured
MGSM (Vals AI)92.8%Not measuredOnly GPT-5 measured
AA Math Index94.318.0GPT-5
AIME 2025 (Artificial Analysis run)94.3%18.0%GPT-5
OTIS Mock AIME91.4%13.8%GPT-5
FrontierMath, tiers 1 to 355.4%Not measuredOnly GPT-5 measured
FrontierMath, tier 422.0%Not measuredOnly GPT-5 measured
ProofBench18.0%Not measuredOnly GPT-5 measured
MATH level 598.1%64.9%GPT-5
LMArena Text: math14071246GPT-5
LMArena Text: mathematical work14051254GPT-5
Neither model has a result on the other 1 board here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
Knowledge benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
MMLU Pro86.5%Not measuredOnly GPT-5 measured
LegalBench (Vals AI)86.0%Not measuredOnly GPT-5 measured
CaseLaw (Vals AI)66.5%Not measuredOnly GPT-5 measured
CorpFin (Vals AI)61.1%Not measuredOnly GPT-5 measured
TaxEval (Vals AI)73.4%Not measuredOnly GPT-5 measured
MortgageTax (Vals AI)65.5%Not measuredOnly GPT-5 measured
MedQA (Vals AI)96.3%Not measuredOnly GPT-5 measured
MedCode (Vals AI)49.6%Not measuredOnly GPT-5 measured
MMLU Pro (Artificial Analysis run)87.1%71.4%GPT-5
SimpleQA Verified50.1%Not measuredOnly GPT-5 measured
MMLUNot measured84.8%Only Phi 4 measured
GeoBench81.0%Not measuredOnly GPT-5 measured
LMArena Text: science14161206GPT-5
LMArena Text: medicine and healthcare14131202GPT-5
LMArena Text: legal and government14311241GPT-5
LMArena Text: business and finance14001199GPT-5
Neither model has a result on the other 1 board here
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Web design benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
Design Arena: websites1185Not measuredOnly GPT-5 measured
Design Arena: UI components1180Not measuredOnly GPT-5 measured
Design Arena: game dev1185Not measuredOnly GPT-5 measured
Design Arena: data viz1221Not measuredOnly GPT-5 measured
Design Arena: 3D1056Not measuredOnly GPT-5 measured
Design Arena: all code categories1168Not measuredOnly GPT-5 measured
Design Arena: SVG1189Not measuredOnly GPT-5 measured
Design Arena: ASCII art1140Not measuredOnly GPT-5 measured
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index190
General benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
SAGE (Vals AI)43.7%Not measuredOnly GPT-5 measured
MedScribe (Vals AI)83.7%Not measuredOnly GPT-5 measured
IFBench73.1%23.5%GPT-5
Long context reasoning (AA-LCR)78.2%0.0%GPT-5
DTBench90.7%Not measuredOnly GPT-5 measured
LMCA40.0%Not measuredOnly GPT-5 measured
Fiction.LiveBench97.2%Not measuredOnly GPT-5 measured
Creative writing (Lech Mazur)86.0%62.6%GPT-5
LMArena Text14061217GPT-5
LMArena Text: instruction following13881201GPT-5
LMArena Text: creative writing13651181GPT-5
LMArena Vision1232Not measuredOnly GPT-5 measured
LMArena Text: English14091231GPT-5
LMArena Text: non-English13971197GPT-5
LMArena Text: Chinese14221212GPT-5
LMArena Text: German14161222GPT-5
LMArena Text: Japanese14091158GPT-5
LMArena Text: Korean13601151GPT-5
LMArena Text: French14091224GPT-5
LMArena Text: Spanish13991234GPT-5
LMArena Text: Russian14071209GPT-5
LMArena Text: Polish1431Not measuredOnly GPT-5 measured
LMArena Text: multi-turn14261206GPT-5
LMArena Text: longer queries13991217GPT-5
LMArena Text: writing and language13841198GPT-5
LMArena Text: entertainment and media13791162GPT-5
LMArena Vision: English1231Not measuredOnly GPT-5 measured
LMArena Vision: Chinese1267Not measuredOnly GPT-5 measured
LMArena Vision: reading text1246Not measuredOnly GPT-5 measured
LMArena Vision: humor1219Not measuredOnly GPT-5 measured
LMArena Vision: captioning1212Not measuredOnly GPT-5 measured
LMArena Vision: recognizing things1263Not measuredOnly GPT-5 measured
LMArena Vision: creative writing1210Not measuredOnly GPT-5 measured
Neither model has a result on the other 13 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-5 and Phi 4
BenchmarkGPT-5Phi 4Leader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

GPT-5$0.0063Fits in one request
Phi 4$0.0001Fits in one request

Phi 4 costs less

Repository review

50K input + 3K output tokens

GPT-5$0.092Fits in one request
Phi 4$0.0039Exceeds the context window

Phi 4 costs less

Agent loop

200K input + 10K output tokens

GPT-5$0.350Fits in one request
Phi 4$0.015Exceeds the context window

Phi 4 costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of GPT-5 and Phi 4
SpecificationOpenAI GPT-5Microsoft Phi 4
MakerOpenAIMicrosoft
WeightsNo public weights listedOpen weights
ReleasedAugust 7, 2025January 10, 2025
Context window400K16K
Input $/Mtok$1.25$0.07
Output $/Mtok$10.00$0.14
Coding index38.8Not benchmarked
Independent measurements behind it30
Cost per solved task$0.39Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-5 better than Phi 4 for coding?

Only GPT-5 has a coding score, 38.8 on the Vibe Coding Index; Phi 4 has none, so the two cannot be ranked on coding. Phi 4 has the edge where output tokens cost 71.4x less.

Which is cheaper, GPT-5 or Phi 4?

GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; Phi 4 costs $0.07 and $0.14. Cost per solved task is known for GPT-5 only.

Which has the bigger context window, GPT-5 or Phi 4?

GPT-5, with 400K tokens against 16K.

How many benchmarks is the GPT-5 vs Phi 4 comparison based on?

44 benchmarks have a result for both models as of October 21, 2026. GPT-5's coding index rests on 3 independent benchmarks and Phi 4's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next