LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

GPT-5 vs GPT-5.4

Updated October 21, 2026. 91 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 GPT-5.4
GPT-5OpenAI
Coding
72.2
Overall
23.0
Output $/M
$10.00
11
GPT-5.4OpenAI
Coding
37.7
Overall
39.0
Output $/M
$15.00
80
vs boards won of 91 shared

Decision reading

Pick GPT-5 for coding: it scores 72.2 on the Vibe Coding Index against 37.7 for GPT-5.4.

Pick GPT-5.4 when the context window is larger, 1.05M against 400K tokens, or when what matters is LMArena WebDev or AA Coding Index, where it leads.

Model AGPT-5OpenAI72.2/1003 independent benchmarks$0.39 per solved taskModel BGPT-5.4OpenAI37.7/1001 independent benchmark$1.02 per solved task
Shared results
91
GPT-5 only
18
GPT-5.4 only
18
Like-for-like categories
8 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 72.2 from 3 benchmarks, GPT-5.4 37.7 from 1 benchmark.
Widest category
Coding, 19.1 points apart on average over 13 shared benchmarks.
Widest single gap
ARC-AGI-2: GPT-5 9.9% against GPT-5.4 74.0%, 64.1 points apart on a 0 to 100 scale.
Output price
GPT-5 $10.00 against GPT-5.4 $15.00 per million output tokens, 1.5x apart.
Context window
GPT-5 takes 400K tokens, GPT-5.4 1.05M.

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-5BarsGPT-5.4Ties
Coding112
Intelligence02
Agentic work12
Reasoning015
Math07
Knowledge49
Web design26
General327
All shared boards1180
CodingLike-for-likeFour of these boards feed the Vibe Coding Index112
Coding benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
Aider polyglot88.0%Not measuredOnly GPT-5 measured
SWE-bench Verified74.4%Not measuredOnly GPT-5 measured
LMArena WebDev14181464GPT-5.4
AA Coding Index37.871.1GPT-5.4
LiveCodeBench (Vals AI)85.9%84.1%GPT-5
SWE-bench Verified (Vals AI)69.0%78.2%GPT-5.4
Vibe Code Bench (Vals AI)20.1%67.4%GPT-5.4
Code migration (Vals AI)Not measured35.0%Only GPT-5.4 measured
ProgramBench (Vals AI)Not measured0.5%Only GPT-5.4 measured
SkillsBench (Vals AI)Not measured51.7%Only GPT-5.4 measured
LiveCodeBench (Artificial Analysis run)84.6%Not measuredOnly GPT-5 measured
Terminal-Bench Hard (Artificial Analysis run)37.9%57.6%GPT-5.4
Terminal-Bench49.6%81.8%GPT-5.4
GSO-Bench6.9%31.4%GPT-5.4
WeirdML60.7%77.7%GPT-5.4
MirrorCodeNot measured15.6%Only GPT-5.4 measured
ALE-Bench807.61607.0GPT-5.4
LMArena Text: coding14361496GPT-5.4
LMArena Text: software and IT14201491GPT-5.4
LMArena WebDev: ReactNot measured1454Only GPT-5.4 measured
LMArena WebDev: HTML14271478GPT-5.4
LMArena WebDev: from an imageNot measured1436Only GPT-5.4 measured
Neither model has a result on the other 16 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
AA Intelligence Index23.039.0GPT-5.4
Epoch Capabilities Index150.0157.3GPT-5.4
Agentic workLike-for-likePublic boards, shown as published, kept out of the index12
Agentic work benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
Design Arena: agents building web appsNot measured1059Only GPT-5.4 measured
Design Arena: agents building full-stack appsNot measured1017Only GPT-5.4 measured
Design Arena: agents building mobile appsNot measured1087Only GPT-5.4 measured
Design Arena: agents building Android appsNot measured1037Only GPT-5.4 measured
Design Arena: agents building Godot gamesNot measured1135Only GPT-5.4 measured
τ²-Bench Telecom86.5%87.1%GPT-5.4
APEX-AgentsNot measured52.4%Only GPT-5.4 measured
METR time horizons69.6%74.3%GPT-5.4
Remote Labor Index1.7%Not measuredOnly GPT-5 measured
GDPval34.8%Not measuredOnly GPT-5 measured
BALROG32.8%Not measuredOnly GPT-5 measured
DeepResearch Bench49.6%35.1%GPT-5
PostTrainBenchNot measured19.0%Only GPT-5.4 measured
Vending-Bench 2Not measured$6,144Only GPT-5.4 measured
Neither model has a result on the other 25 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index015
Reasoning benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
GPQA Diamond85.6%91.7%GPT-5.4
MMMU Pro (Vals AI)81.5%87.5%GPT-5.4
GPQA Diamond (Artificial Analysis run)85.4%92.0%GPT-5.4
Humanity's Last Exam (Artificial Analysis run)28.5%43.7%GPT-5.4
GPQA Diamond (Epoch AI run)86.2%93.3%GPT-5.4
ARC-AGI65.7%93.7%GPT-5.4
ARC-AGI-29.9%74.0%GPT-5.4
SimpleBench56.7%Not measuredOnly GPT-5 measured
Chess puzzles36.0%48.0%GPT-5.4
Mystery game puzzles23.0%37.0%GPT-5.4
EBR-bench12.7%25.4%GPT-5.4
VPCT66.0%Not measuredOnly GPT-5 measured
LMArena Text: hard prompts14151485GPT-5.4
LMArena Text: expert14191507GPT-5.4
LMArena Text: hard prompts in English14191479GPT-5.4
LMArena Vision: diagrams12521327GPT-5.4
LMArena Vision: homework12781336GPT-5.4
Neither model has a result on the other 4 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index07
Math benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
AIME93.4%96.7%GPT-5.4
MATH 500 (Vals AI)96.0%Not measuredOnly GPT-5 measured
MGSM (Vals AI)92.8%Not measuredOnly GPT-5 measured
AA Math Index94.3Not measuredOnly GPT-5 measured
AIME 2025 (Artificial Analysis run)94.3%Not measuredOnly GPT-5 measured
OTIS Mock AIME91.4%97.8%GPT-5.4
FrontierMath, tiers 1 to 355.4%78.6%GPT-5.4
FrontierMath, tier 422.0%49.0%GPT-5.4
ProofBench18.0%56.0%GPT-5.4
MATH level 598.1%Not measuredOnly GPT-5 measured
LMArena Text: math14071488GPT-5.4
LMArena Text: mathematical work14051497GPT-5.4
Neither model has a result on the other 1 board here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index49
Knowledge benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
MMLU Pro86.5%87.5%GPT-5.4
LegalBench (Vals AI)86.0%86.0%GPT-5.4
CaseLaw (Vals AI)66.5%63.8%GPT-5
CorpFin (Vals AI)61.1%65.3%GPT-5.4
TaxEval (Vals AI)73.4%74.0%GPT-5.4
MortgageTax (Vals AI)65.5%68.3%GPT-5.4
MedQA (Vals AI)96.3%96.1%GPT-5
MedCode (Vals AI)49.6%41.3%GPT-5
MMLU Pro (Artificial Analysis run)87.1%Not measuredOnly GPT-5 measured
SimpleQA Verified50.1%45.1%GPT-5
GeoBench81.0%Not measuredOnly GPT-5 measured
LMArena Text: science14161479GPT-5.4
LMArena Text: medicine and healthcare14131458GPT-5.4
LMArena Text: legal and government14311486GPT-5.4
LMArena Text: business and finance14001474GPT-5.4
Neither model has a result on the other 2 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index26
Web design benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
Design Arena: websites11851199GPT-5.4
Design Arena: UI components11791215GPT-5.4
Design Arena: game dev11851195GPT-5.4
Design Arena: data viz12211196GPT-5
Design Arena: 3D10561065GPT-5.4
Design Arena: all code categories11681180GPT-5.4
Design Arena: SVG11881141GPT-5
Design Arena: ASCII art11401188GPT-5.4
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index327
General benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
SAGE (Vals AI)43.7%43.3%GPT-5
MedScribe (Vals AI)83.7%77.5%GPT-5
IFBench73.1%73.9%GPT-5.4
Long context reasoning (AA-LCR)78.2%82.0%GPT-5.4
DTBench90.7%94.4%GPT-5.4
LMCA40.0%52.0%GPT-5.4
Fiction.LiveBench97.2%Not measuredOnly GPT-5 measured
Creative writing (Lech Mazur)86.0%Not measuredOnly GPT-5 measured
CL-benchNot measured27.9%Only GPT-5.4 measured
Furniture assemblyNot measured37.5%Only GPT-5.4 measured
CL-bench LifeNot measured21.7%Only GPT-5.4 measured
LMArena Text14061469GPT-5.4
LMArena Text: instruction following13881469GPT-5.4
LMArena Text: creative writing13651439GPT-5.4
LMArena Vision12321303GPT-5.4
LMArena Text: English14091467GPT-5.4
LMArena Text: non-English13971465GPT-5.4
LMArena Text: Chinese14221520GPT-5.4
LMArena Text: German14161472GPT-5.4
LMArena Text: Japanese14091485GPT-5.4
LMArena Text: Korean13601449GPT-5.4
LMArena Text: French14091493GPT-5.4
LMArena Text: Spanish13991452GPT-5.4
LMArena Text: Russian14071480GPT-5.4
LMArena Text: Polish14311472GPT-5.4
LMArena Text: multi-turn14261482GPT-5.4
LMArena Text: longer queries13991473GPT-5.4
LMArena Text: writing and language13841458GPT-5.4
LMArena Text: entertainment and media13791438GPT-5.4
LMArena Vision: English12311301GPT-5.4
LMArena Vision: Chinese12671354GPT-5.4
LMArena Vision: reading text12461312GPT-5.4
LMArena Vision: humor12191284GPT-5.4
LMArena Vision: captioning1212Not measuredOnly GPT-5 measured
LMArena Vision: recognizing things12631251GPT-5
LMArena Vision: creative writing12101283GPT-5.4
LMArena DocumentNot measured1471Only GPT-5.4 measured
Neither model has a result on the other 9 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-5 and GPT-5.4
BenchmarkGPT-5GPT-5.4Leader
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
GPT-5.4$0.010Fits in one request

GPT-5 costs less

Repository review

50K input + 3K output tokens

GPT-5$0.092Fits in one request
GPT-5.4$0.170Fits in one request

GPT-5 costs less

Agent loop

200K input + 10K output tokens

GPT-5$0.350Fits in one request
GPT-5.4$0.650Fits in one request

GPT-5 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 GPT-5.4
SpecificationOpenAI GPT-5OpenAI GPT-5.4
MakerOpenAIOpenAI
WeightsNo public weights listedNo public weights listed
ReleasedAugust 7, 2025March 5, 2026
Context window400K1.05M
Input $/Mtok$1.25$2.50
Output $/Mtok$10.00$15.00
Coding index72.237.7
Independent measurements behind it31
Cost per solved task$0.39$1.02
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-5 better than GPT-5.4 for coding?

Pick GPT-5 for coding: it scores 72.2 on the Vibe Coding Index against 37.7 for GPT-5.4. Pick GPT-5.4 when the context window is larger, 1.05M against 400K tokens, or when what matters is LMArena WebDev or AA Coding Index, where it leads.

Which is cheaper, GPT-5 or GPT-5.4?

GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; GPT-5.4 costs $2.50 and $15.00. Per solved SWE-bench Verified task, GPT-5 costs $0.39 and GPT-5.4 $1.02.

Which has the bigger context window, GPT-5 or GPT-5.4?

GPT-5.4, with 1.05M tokens against 400K.

How many benchmarks is the GPT-5 vs GPT-5.4 comparison based on?

91 benchmarks have a result for both models as of October 21, 2026. GPT-5's coding index rests on 3 independent benchmarks and GPT-5.4's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next