Enter opens the highlighted result. Esc closes.

GPT-4 Turbo vs GPT-5.3-Codex

GPT-5.3-Codex leads GPT-4 Turbo on the Artificial Analysis Intelligence Index, 32.5 to 7.0, and GPT-5.3-Codex costs 2.1x less per output token ($14.00 vs $30.00 per million). Updated October 21, 2026; 3 benchmarks have a result for both.

Replace GPT-4 Turbo Replace GPT-5.3-Codex
GPT-4 TurboOpenAI
Coding
No data
Overall
No data
Output $/M
$30.00
0
GPT-5.3-CodexOpenAI
Coding
60.0
Overall
No data
Output $/M
$14.00
3
vs boards won of 3 shared

GPT-4 Turbo or GPT-5.3-Codex: which to pick

Only GPT-5.3-Codex has a coding score, 60.0 on the Vibe Coding Index; GPT-4 Turbo has none, so the two cannot be ranked on coding.

Compare them on price, context and the boards below.

Model AGPT-4 TurboOpenAINot benchmarkedNo coding benchmarkNo cost per solved taskModel BGPT-5.3-CodexOpenAI60.0/1003 independent benchmarks$0.59 per solved task
Shared results
3
GPT-4 Turbo only
39
GPT-5.3-Codex only
31
Like-for-like categories
2 of 9

Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.

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-4 Turbo not benchmarked, GPT-5.3-Codex 60.0 from 3 benchmarks.
Widest category
Reasoning, 39.4 points apart on average over 1 shared benchmark.
Widest single gap
Humanity's Last Exam (Artificial Analysis run): GPT-4 Turbo 3.1% against GPT-5.3-Codex 42.5%, 39.4 points apart on a 0 to 100 scale.
Output price
GPT-4 Turbo $30.00 against GPT-5.3-Codex $14.00 per million output tokens, 2.1x apart.
Context window
GPT-4 Turbo takes 128K tokens, GPT-5.3-Codex 400K.

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-4 TurboBarsGPT-5.3-CodexTies
Intelligence02
Reasoning01
All shared boards03
CodingDirectional onlySix of these boards feed the Vibe Coding Index16 boards, none shared
Coding benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
LMArena WebDevNot measured1407Only GPT-5.3-Codex measured
AA Coding Index21.5Not measuredOnly GPT-4 Turbo measured
LiveCodeBench (Vals AI)Not measured87.3%Only GPT-5.3-Codex measured
SWE-bench Verified (Vals AI)Not measured78.0%Only GPT-5.3-Codex measured
Vibe Code Bench (Vals AI)Not measured61.8%Only GPT-5.3-Codex measured
IOI (Vals AI)Not measured53.8%Only GPT-5.3-Codex measured
LiveCodeBench (Artificial Analysis run)29.1%Not measuredOnly GPT-4 Turbo measured
Terminal-Bench Hard (Artificial Analysis run)Not measured53.0%Only GPT-5.3-Codex measured
Terminal-BenchNot measured78.4%Only GPT-5.3-Codex measured
WeirdMLNot measured79.3%Only GPT-5.3-Codex measured
ALE-BenchNot measured1655.2Only GPT-5.3-Codex measured
LMArena Text: coding1268Not measuredOnly GPT-4 Turbo measured
LMArena Text: software and IT1267Not measuredOnly GPT-4 Turbo measured
LMArena WebDev: ReactNot measured1395Only GPT-5.3-Codex measured
LMArena WebDev: HTMLNot measured1422Only GPT-5.3-Codex measured
LMArena WebDev: from an imageNot measured1443Only GPT-5.3-Codex measured
Neither model has a result on the other 23 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
AA Intelligence Index7.032.5GPT-5.3-Codex
Epoch Capabilities Index127.3157.1GPT-5.3-Codex
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Agentic work benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
Design Arena: agents building web appsNot measured1032Only GPT-5.3-Codex measured
Design Arena: agents building full-stack appsNot measured983Only GPT-5.3-Codex measured
Design Arena: agents building mobile appsNot measured1054Only GPT-5.3-Codex measured
Design Arena: agents building Android appsNot measured1084Only GPT-5.3-Codex measured
Design Arena: agents building Godot gamesNot measured1124Only GPT-5.3-Codex measured
τ²-Bench TelecomNot measured86.0%Only GPT-5.3-Codex measured
METR time horizonsNot measured74.5%Only GPT-5.3-Codex measured
Vending-Bench 2Not measured$5,940Only GPT-5.3-Codex measured
Neither model has a result on the other 33 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
Reasoning benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
GPQA Diamond (Artificial Analysis run)Not measured91.5%Only GPT-5.3-Codex measured
Humanity's Last Exam (Artificial Analysis run)3.1%42.5%GPT-5.3-Codex
LMArena Text: hard prompts1251Not measuredOnly GPT-4 Turbo measured
LMArena Text: expert1223Not measuredOnly GPT-4 Turbo measured
LMArena Text: hard prompts in English1266Not measuredOnly GPT-4 Turbo measured
Neither model has a result on the other 17 boards here
MathDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
Math benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
LMArena Text: math1272Not measuredOnly GPT-4 Turbo measured
LMArena Text: mathematical work1268Not measuredOnly GPT-4 Turbo measured
Neither model has a result on the other 11 boards here
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Knowledge benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
LegalBench (Vals AI)80.5%Not measuredOnly GPT-4 Turbo measured
MedQA (Vals AI)82.0%Not measuredOnly GPT-4 Turbo measured
MMLU Pro (Artificial Analysis run)69.4%Not measuredOnly GPT-4 Turbo measured
MMLU81.3%Not measuredOnly GPT-4 Turbo measured
LMArena Text: science1256Not measuredOnly GPT-4 Turbo measured
LMArena Text: medicine and healthcare1239Not measuredOnly GPT-4 Turbo measured
LMArena Text: legal and government1265Not measuredOnly GPT-4 Turbo measured
LMArena Text: business and finance1222Not measuredOnly GPT-4 Turbo measured
Neither model has a result on the other 10 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Web design benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
Design Arena: websitesNot measured1163Only GPT-5.3-Codex measured
Design Arena: UI componentsNot measured1137Only GPT-5.3-Codex measured
Design Arena: game devNot measured1161Only GPT-5.3-Codex measured
Design Arena: data vizNot measured1164Only GPT-5.3-Codex measured
Design Arena: 3DNot measured1008Only GPT-5.3-Codex measured
Design Arena: all code categoriesNot measured1145Only GPT-5.3-Codex measured
Design Arena: SVGNot measured1129Only GPT-5.3-Codex measured
Design Arena: ASCII artNot measured1152Only GPT-5.3-Codex measured
Neither model has a result on the other 1 board here
GeneralDirectional onlyPublic boards, shown as published, kept out of the index24 boards, none shared
General benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
IFBenchNot measured75.4%Only GPT-5.3-Codex measured
Long context reasoning (AA-LCR)Not measured83.3%Only GPT-5.3-Codex measured
DTBench61.6%Not measuredOnly GPT-4 Turbo measured
LMCA9.8%Not measuredOnly GPT-4 Turbo measured
LMArena Text1272Not measuredOnly GPT-4 Turbo measured
LMArena Text: instruction following1249Not measuredOnly GPT-4 Turbo measured
LMArena Text: creative writing1268Not measuredOnly GPT-4 Turbo measured
LMArena Vision1090Not measuredOnly GPT-4 Turbo measured
LMArena Text: English1291Not measuredOnly GPT-4 Turbo measured
LMArena Text: non-English1245Not measuredOnly GPT-4 Turbo measured
LMArena Text: Chinese1242Not measuredOnly GPT-4 Turbo measured
LMArena Text: German1259Not measuredOnly GPT-4 Turbo measured
LMArena Text: Japanese1194Not measuredOnly GPT-4 Turbo measured
LMArena Text: Korean1187Not measuredOnly GPT-4 Turbo measured
LMArena Text: French1276Not measuredOnly GPT-4 Turbo measured
LMArena Text: Spanish1260Not measuredOnly GPT-4 Turbo measured
LMArena Text: Russian1259Not measuredOnly GPT-4 Turbo measured
LMArena Text: Polish1302Not measuredOnly GPT-4 Turbo measured
LMArena Text: multi-turn1267Not measuredOnly GPT-4 Turbo measured
LMArena Text: longer queries1254Not measuredOnly GPT-4 Turbo measured
LMArena Text: writing and language1279Not measuredOnly GPT-4 Turbo measured
LMArena Text: entertainment and media1258Not measuredOnly GPT-4 Turbo measured
LMArena Vision: English1097Not measuredOnly GPT-4 Turbo measured
LMArena Vision: Chinese1045Not measuredOnly GPT-4 Turbo measured
Neither model has a result on the other 22 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-4 Turbo and GPT-5.3-Codex
BenchmarkGPT-4 TurboGPT-5.3-CodexLeader
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-4 Turbo$0.025Fits in one request
GPT-5.3-Codex$0.0088Fits in one request

GPT-5.3-Codex costs less

Repository review

50K input + 3K output tokens

GPT-4 Turbo$0.590Fits in one request
GPT-5.3-Codex$0.130Fits in one request

GPT-5.3-Codex costs less

Agent loop

200K input + 10K output tokens

GPT-4 Turbo$2.30Exceeds the context window
GPT-5.3-Codex$0.490Fits in one request

GPT-5.3-Codex 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-4 Turbo and GPT-5.3-Codex
SpecificationOpenAI GPT-4 TurboOpenAI GPT-5.3-Codex
MakerOpenAIOpenAI
WeightsNo public weights listedNo public weights listed
ReleasedApril 9, 2024February 24, 2026
Context window128K400K
Input $/Mtok$10.00$1.75
Output $/Mtok$30.00$14.00
Coding indexNot benchmarked60.0
Independent measurements behind it03
Cost per solved taskNot computable$0.59
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-4 Turbo better than GPT-5.3-Codex for coding?

Only GPT-5.3-Codex has a coding score, 60.0 on the Vibe Coding Index; GPT-4 Turbo has none, so the two cannot be ranked on coding. Compare them on price, context and the boards below.

Which is cheaper, GPT-4 Turbo or GPT-5.3-Codex?

GPT-4 Turbo costs $10.00 per million input tokens and $30.00 per million output tokens; GPT-5.3-Codex costs $1.75 and $14.00. Cost per solved task is known for GPT-5.3-Codex only.

Which has the bigger context window, GPT-4 Turbo or GPT-5.3-Codex?

GPT-5.3-Codex, with 400K tokens against 128K.

How many benchmarks is the GPT-4 Turbo vs GPT-5.3-Codex comparison based on?

3 benchmarks have a result for both models as of October 21, 2026. GPT-4 Turbo's coding index rests on 0 independent benchmarks and GPT-5.3-Codex's on 3. Where only one model has a result, the page says so instead of naming a winner.

Compare next