Enter opens the highlighted result. Esc closes.

Gemma 3 4B vs GPT-5

GPT-5 leads Gemma 3 4B on the Artificial Analysis Intelligence Index, 23.0 to 4.8, and Gemma 3 4B costs 133.3x less per output token ($0.075 vs $10.00 per million). Updated October 21, 2026; 39 benchmarks have a result for both.

Replace Gemma 3 4B Replace GPT-5
Gemma 3 4BGoogle
Coding
No data
Overall
4.8
Output $/M
$0.075
0
GPT-5OpenAI
Coding
38.8
Overall
23.0
Output $/M
$10.00
39
vs boards won of 39 shared

Gemma 3 4B or GPT-5: which to pick

Only GPT-5 has a coding score, 38.8 on the Vibe Coding Index; Gemma 3 4B has none, so the two cannot be ranked on coding.

Pick Gemma 3 4B for output tokens 133.3x cheaper ($0.075 vs $10.00 per million).

Model AGemma 3 4BGoogleNot benchmarkedNo coding benchmarkNo cost per solved taskModel BGPT-5OpenAI38.8/1003 independent benchmarks$0.39 per solved task
Shared results
39
Gemma 3 4B only
0
GPT-5 only
70
Like-for-like categories
7 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
Gemma 3 4B not benchmarked, GPT-5 38.8 from 3 benchmarks.
Widest category
Agentic work, 81.6 points apart on average over 1 shared benchmark.
Widest single gap
OTIS Mock AIME: Gemma 3 4B 7.5% against GPT-5 91.4%, 83.9 points apart on a 0 to 100 scale.
Output price
Gemma 3 4B $0.075 against GPT-5 $10.00 per million output tokens, 133.3x apart.
Context window
Gemma 3 4B takes 131K tokens, GPT-5 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
CategoryGemma 3 4BBarsGPT-5Ties
Coding05
Intelligence02
Agentic work01
Reasoning06
Math05
Knowledge05
General015
All shared boards039
CodingLike-for-likeSix of these boards feed the Vibe Coding Index05
Coding benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
Aider polyglotNot measured88.0%Only GPT-5 measured
SWE-bench VerifiedNot measured74.4%Only GPT-5 measured
LMArena WebDevNot measured1418Only GPT-5 measured
AA Coding Index2.737.8GPT-5
LiveCodeBench (Vals AI)Not measured85.9%Only GPT-5 measured
SWE-bench Verified (Vals AI)Not measured69.0%Only GPT-5 measured
Vibe Code Bench (Vals AI)Not measured20.1%Only GPT-5 measured
LiveCodeBench (Artificial Analysis run)11.2%84.6%GPT-5
Terminal-Bench Hard (Artificial Analysis run)0.8%37.9%GPT-5
Terminal-BenchNot measured49.6%Only GPT-5 measured
GSO-BenchNot measured6.9%Only GPT-5 measured
WeirdMLNot measured60.7%Only GPT-5 measured
ALE-BenchNot measured807.6Only GPT-5 measured
LMArena Text: coding12301436GPT-5
LMArena Text: software and IT12691420GPT-5
LMArena WebDev: HTMLNot measured1427Only 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 index02
Intelligence benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
AA Intelligence Index4.823.0GPT-5
Epoch Capabilities Index115.9150.0GPT-5
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
Agentic work benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
τ²-Bench Telecom5.0%86.5%GPT-5
METR time horizonsNot measured69.6%Only GPT-5 measured
Remote Labor IndexNot measured1.7%Only GPT-5 measured
GDPvalNot measured34.8%Only GPT-5 measured
BALROGNot measured32.8%Only GPT-5 measured
DeepResearch BenchNot measured49.6%Only 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 index06
Reasoning benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
GPQA DiamondNot measured85.6%Only GPT-5 measured
MMMU Pro (Vals AI)Not measured81.5%Only GPT-5 measured
GPQA Diamond (Artificial Analysis run)29.1%85.4%GPT-5
Humanity's Last Exam (Artificial Analysis run)5.3%28.5%GPT-5
GPQA Diamond (Epoch AI run)23.2%86.2%GPT-5
ARC-AGINot measured65.7%Only GPT-5 measured
ARC-AGI-2Not measured9.9%Only GPT-5 measured
SimpleBenchNot measured56.7%Only GPT-5 measured
Chess puzzlesNot measured36.0%Only GPT-5 measured
Mystery game puzzlesNot measured23.0%Only GPT-5 measured
EBR-benchNot measured12.7%Only GPT-5 measured
VPCTNot measured66.0%Only GPT-5 measured
LMArena Text: hard prompts12531415GPT-5
LMArena Text: expert12231419GPT-5
LMArena Text: hard prompts in English12531419GPT-5
LMArena Vision: diagramsNot measured1252Only GPT-5 measured
LMArena Vision: homeworkNot measured1278Only 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 index05
Math benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
AIMENot measured93.4%Only GPT-5 measured
MATH 500 (Vals AI)Not measured96.0%Only GPT-5 measured
MGSM (Vals AI)Not measured92.8%Only GPT-5 measured
AA Math Index12.794.3GPT-5
AIME 2025 (Artificial Analysis run)12.7%94.3%GPT-5
OTIS Mock AIME7.5%91.4%GPT-5
FrontierMath, tiers 1 to 3Not measured55.4%Only GPT-5 measured
FrontierMath, tier 4Not measured22.0%Only GPT-5 measured
ProofBenchNot measured18.0%Only GPT-5 measured
MATH level 5Not measured98.1%Only GPT-5 measured
LMArena Text: math12391407GPT-5
LMArena Text: mathematical work12561405GPT-5
Neither model has a result on the other 1 board here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index05
Knowledge benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
MMLU ProNot measured86.5%Only GPT-5 measured
LegalBench (Vals AI)Not measured86.0%Only GPT-5 measured
CaseLaw (Vals AI)Not measured66.5%Only GPT-5 measured
CorpFin (Vals AI)Not measured61.1%Only GPT-5 measured
TaxEval (Vals AI)Not measured73.4%Only GPT-5 measured
MortgageTax (Vals AI)Not measured65.5%Only GPT-5 measured
MedQA (Vals AI)Not measured96.3%Only GPT-5 measured
MedCode (Vals AI)Not measured49.6%Only GPT-5 measured
MMLU Pro (Artificial Analysis run)41.7%87.1%GPT-5
SimpleQA VerifiedNot measured50.1%Only GPT-5 measured
GeoBenchNot measured81.0%Only GPT-5 measured
LMArena Text: science13061416GPT-5
LMArena Text: medicine and healthcare13421413GPT-5
LMArena Text: legal and government13271431GPT-5
LMArena Text: business and finance13111400GPT-5
Neither model has a result on the other 2 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
Web design benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
Design Arena: websitesNot measured1184Only GPT-5 measured
Design Arena: UI componentsNot measured1180Only GPT-5 measured
Design Arena: game devNot measured1185Only GPT-5 measured
Design Arena: data vizNot measured1221Only GPT-5 measured
Design Arena: 3DNot measured1056Only GPT-5 measured
Design Arena: all code categoriesNot measured1168Only GPT-5 measured
Design Arena: SVGNot measured1189Only GPT-5 measured
Design Arena: ASCII artNot measured1141Only 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 index015
General benchmarks for Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
SAGE (Vals AI)Not measured43.7%Only GPT-5 measured
MedScribe (Vals AI)Not measured83.7%Only GPT-5 measured
IFBench28.3%73.1%GPT-5
Long context reasoning (AA-LCR)6.7%78.2%GPT-5
DTBench50.9%90.7%GPT-5
LMCA2.8%40.0%GPT-5
Fiction.LiveBenchNot measured97.2%Only GPT-5 measured
Creative writing (Lech Mazur)Not measured86.0%Only GPT-5 measured
LMArena Text12911406GPT-5
LMArena Text: instruction following12391388GPT-5
LMArena Text: creative writing12701365GPT-5
LMArena VisionNot measured1232Only GPT-5 measured
LMArena Text: English13001409GPT-5
LMArena Text: non-English12731397GPT-5
LMArena Text: ChineseNot measured1422Only GPT-5 measured
LMArena Text: German12811416GPT-5
LMArena Text: JapaneseNot measured1409Only GPT-5 measured
LMArena Text: KoreanNot measured1360Only GPT-5 measured
LMArena Text: FrenchNot measured1409Only GPT-5 measured
LMArena Text: SpanishNot measured1399Only GPT-5 measured
LMArena Text: Russian12941407GPT-5
LMArena Text: PolishNot measured1431Only GPT-5 measured
LMArena Text: multi-turn12551426GPT-5
LMArena Text: longer queries12731399GPT-5
LMArena Text: writing and language12881384GPT-5
LMArena Text: entertainment and media12331379GPT-5
LMArena Vision: EnglishNot measured1231Only GPT-5 measured
LMArena Vision: ChineseNot measured1267Only GPT-5 measured
LMArena Vision: reading textNot measured1246Only GPT-5 measured
LMArena Vision: humorNot measured1219Only GPT-5 measured
LMArena Vision: captioningNot measured1212Only GPT-5 measured
LMArena Vision: recognizing thingsNot measured1263Only GPT-5 measured
LMArena Vision: creative writingNot measured1210Only 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 Gemma 3 4B and GPT-5
BenchmarkGemma 3 4BGPT-5Leader
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

Gemma 3 4B$0.0001Fits in one request
GPT-5$0.0063Fits in one request

Gemma 3 4B costs less

Repository review

50K input + 3K output tokens

Gemma 3 4B$0.0025Fits in one request
GPT-5$0.092Fits in one request

Gemma 3 4B costs less

Agent loop

200K input + 10K output tokens

Gemma 3 4B$0.0097Exceeds the context window
GPT-5$0.350Fits in one request

Gemma 3 4B 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 Gemma 3 4B and GPT-5
SpecificationGoogle Gemma 3 4BOpenAI GPT-5
MakerGoogleOpenAI
WeightsOpen weightsNo public weights listed
ReleasedMarch 13, 2025August 7, 2025
Context window131K400K
Input $/Mtok$0.045$1.25
Output $/Mtok$0.075$10.00
Coding indexNot benchmarked38.8
Independent measurements behind it03
Cost per solved taskNot computable$0.39
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemma 3 4B better than GPT-5 for coding?

Only GPT-5 has a coding score, 38.8 on the Vibe Coding Index; Gemma 3 4B has none, so the two cannot be ranked on coding. Pick Gemma 3 4B for output tokens 133.3x cheaper ($0.075 vs $10.00 per million).

Which is cheaper, Gemma 3 4B or GPT-5?

Gemma 3 4B costs $0.045 per million input tokens and $0.075 per million output tokens; GPT-5 costs $1.25 and $10.00. Cost per solved task is known for GPT-5 only.

Which has the bigger context window, Gemma 3 4B or GPT-5?

GPT-5, with 400K tokens against 131K.

How many benchmarks is the Gemma 3 4B vs GPT-5 comparison based on?

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

Compare next