Enter opens the highlighted result. Esc closes.

GPT-4 vs o1

Updated October 21, 2026. 35 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace GPT-4 Replace o1
GPT-4OpenAI
Coding
22.4
Overall
6.7
Output $/M
$60.00
0
o1OpenAI
Coding
61.7
Overall
15.2
Output $/M
$60.00
35
vs boards won of 35 shared

Decision reading

Pick o1 for coding: it scores 61.7 on the Vibe Coding Index against 22.4 for GPT-4.

GPT-4 is neither cheaper nor larger in context, and leads no benchmark both models were measured on.

Model AGPT-4OpenAI22.4/1001 independent benchmarkNo cost per solved taskModel Bo1OpenAI61.7/1001 independent benchmarkNo cost per solved task
Shared results
35
GPT-4 only
10
o1 only
33
Like-for-like categories
6 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-4 22.4 from 1 benchmark, o1 61.7 from 1 benchmark.
Widest category
Reasoning, 32.8 points apart on average over 4 shared benchmarks.
Widest single gap
LMArena Text: Korean: GPT-4 1087 against o1 1396, 51.5 points apart on a 0 to 100 scale.
Output price
Both cost $60.00 per million output tokens.
Context window
GPT-4 takes 8K tokens, o1 200K.

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-4Barso1Ties
Coding03
Intelligence02
Reasoning04
Math02
Knowledge05
General019
All shared boards035
CodingLike-for-likeFour of these boards feed the Vibe Coding Index03
Coding benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
Aider polyglotNot measured61.7%Only o1 measured
SWE-bench Verified22.4%Not measuredOnly GPT-4 measured
SWE-bench Lite39.7%Not measuredOnly GPT-4 measured
SWE-bench (full test set)12.5%Not measuredOnly GPT-4 measured
AA Coding Index13.139.7o1
LiveCodeBench (Vals AI)Not measured50.3%Only o1 measured
LiveCodeBench (Artificial Analysis run)Not measured67.9%Only o1 measured
Terminal-Bench Hard (Artificial Analysis run)Not measured12.9%Only o1 measured
WeirdMLNot measured46.1%Only o1 measured
LMArena Text: coding12091367o1
LMArena Text: software and IT11951366o1
Neither model has a result on the other 27 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
AA Intelligence Index6.715.2o1
Epoch Capabilities Index125.7141.7o1
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
Agentic work benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
τ²-Bench TelecomNot measured62.6%Only o1 measured
METR time horizonsNot measured51.1%Only o1 measured
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index04
Reasoning benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
GPQA DiamondNot measured73.2%Only o1 measured
MMMU Pro (Vals AI)Not measured77.4%Only o1 measured
GPQA Diamond (Artificial Analysis run)34.9%74.7%o1
Humanity's Last Exam (Artificial Analysis run)Not measured7.0%Only o1 measured
GPQA Diamond (Epoch AI run)Not measured76.8%Only o1 measured
ARC-AGINot measured30.7%Only o1 measured
SimpleBenchNot measured40.1%Only o1 measured
Chess puzzlesNot measured10.0%Only o1 measured
VPCTNot measured37.0%Only o1 measured
BIG-Bench Hard75.1%Not measuredOnly GPT-4 measured
LMArena Text: hard prompts12001371o1
LMArena Text: expert11491361o1
LMArena Text: hard prompts in English12091372o1
Neither model has a result on the other 8 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index02
Math benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
AIMENot measured71.5%Only o1 measured
MATH 500 (Vals AI)Not measured90.4%Only o1 measured
MGSM (Vals AI)Not measured89.3%Only o1 measured
OTIS Mock AIMENot measured73.3%Only o1 measured
FrontierMath, tiers 1 to 3Not measured14.7%Only o1 measured
MATH level 5Not measured94.7%Only o1 measured
GSM8K92.0%Not measuredOnly GPT-4 measured
LMArena Text: math12301388o1
LMArena Text: mathematical work12281390o1
Neither model has a result on the other 4 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index05
Knowledge benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
MMLU ProNot measured83.5%Only o1 measured
LegalBench (Vals AI)Not measured80.4%Only o1 measured
TaxEval (Vals AI)Not measured74.3%Only o1 measured
MedQA (Vals AI)Not measured96.5%Only o1 measured
MMLU Pro (Artificial Analysis run)56.2%84.1%o1
SimpleQA VerifiedNot measured41.1%Only o1 measured
MMLU86.4%Not measuredOnly GPT-4 measured
GeoBenchNot measured80.0%Only o1 measured
AI2 Reasoning Challenge96.3%Not measuredOnly GPT-4 measured
LMArena Text: science11771369o1
LMArena Text: medicine and healthcare11451355o1
LMArena Text: legal and government11911368o1
LMArena Text: business and finance11451328o1
Neither model has a result on the other 4 boards here
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Web design benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index019
General benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
IFBench33.2%70.3%o1
Long context reasoning (AA-LCR)Not measured65.0%Only o1 measured
DTBench62.7%74.7%o1
LMCA17.1%22.3%o1
Fiction.LiveBenchNot measured83.3%Only o1 measured
Creative writing (Lech Mazur)Not measured70.2%Only o1 measured
HellaSwag95.3%Not measuredOnly GPT-4 measured
WinoGrande87.5%Not measuredOnly GPT-4 measured
LMArena Text12061366o1
LMArena Text: instruction following12011367o1
LMArena Text: creative writing11911347o1
LMArena VisionNot measured1168Only o1 measured
LMArena Text: English12201372o1
LMArena Text: non-English11941358o1
LMArena Text: Chinese11841394o1
LMArena Text: German11981337o1
LMArena Text: Japanese11371346o1
LMArena Text: Korean10871396o1
LMArena Text: French12201344o1
LMArena Text: Spanish11971345o1
LMArena Text: Russian11951356o1
LMArena Text: Polish1265Not measuredOnly GPT-4 measured
LMArena Text: multi-turn12061355o1
LMArena Text: longer queries11901378o1
LMArena Text: writing and language12191360o1
LMArena Text: entertainment and media11841358o1
LMArena Vision: EnglishNot measured1162Only o1 measured
LMArena Vision: ChineseNot measured1235Only o1 measured
Neither model has a result on the other 18 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-4 and o1
BenchmarkGPT-4o1Leader
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$0.060Fits in one request
o1$0.045Fits in one request

o1 costs less

Repository review

50K input + 3K output tokens

GPT-4$1.68Exceeds the context window
o1$0.930Fits in one request

o1 costs less

Agent loop

200K input + 10K output tokens

GPT-4$6.60Exceeds the context window
o1$3.60Exceeds the context window

o1 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 and o1
SpecificationOpenAI GPT-4OpenAI o1
MakerOpenAIOpenAI
WeightsNo public weights listedNo public weights listed
ReleasedMay 28, 2023December 17, 2024
Context window8K200K
Input $/Mtok$30.00$15.00
Output $/Mtok$60.00$60.00
Coding index22.461.7
Independent measurements behind it11
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-4 better than o1 for coding?

Pick o1 for coding: it scores 61.7 on the Vibe Coding Index against 22.4 for GPT-4. GPT-4 is neither cheaper nor larger in context, and leads no benchmark both models were measured on.

Which is cheaper, GPT-4 or o1?

GPT-4 costs $30.00 per million input tokens and $60.00 per million output tokens; o1 costs $15.00 and $60.00. Cost per solved task is known for neither of them.

Which has the bigger context window, GPT-4 or o1?

o1, with 200K tokens against 8K.

How many benchmarks is the GPT-4 vs o1 comparison based on?

35 benchmarks have a result for both models as of October 21, 2026. GPT-4's coding index rests on 1 independent benchmark and o1's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next