Enter opens the highlighted result. Esc closes.

Gemma 3 12B vs Gemma 4 26B A4B

Gemma 4 26B A4B leads Gemma 3 12B on the Artificial Analysis Intelligence Index, 16.7 to 3.8, and Gemma 3 12B costs 3.7x less per output token ($0.1625 vs $0.60 per million). Updated October 22, 2026; 36 benchmarks have a result for both.

Replace Gemma 3 12B Replace Gemma 4 26B A4B
Gemma 3 12BGoogle
Coding
No data
Overall
28.0
Output $/M
$0.1625
0
Gemma 4 26B A4BGoogle
Coding
No data
Overall
54.7
Output $/M
$0.60
36
vs boards won of 36 shared

Gemma 3 12B or Gemma 4 26B A4B: which to pick

Neither model has a coding score yet, so the two cannot be ranked on coding.

Pick Gemma 3 12B for output tokens 3.7x cheaper ($0.1625 vs $0.60 per million) and work measured by AA Coding Index or SciCode, where it leads.

Model AGemma 3 12BGoogleNot benchmarkedNo coding benchmarkNo cost per solved taskModel BGemma 4 26B A4BGoogleNot benchmarked1 independent benchmarkNo cost per solved task
Shared results
36
Gemma 3 12B only
7
Gemma 4 26B A4B only
19
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 12B not benchmarked, Gemma 4 26B A4B not benchmarked.
Widest category
Math, 38.6 points apart on average over 3 shared benchmarks.
Widest single gap
OTIS Mock AIME: Gemma 3 12B 16.7% against Gemma 4 26B A4B 82.2%, 65.6 points apart on a 0 to 100 scale.
Output price
Gemma 3 12B $0.1625 against Gemma 4 26B A4B $0.60 per million output tokens, 3.7x apart.
Context window
Gemma 3 12B takes 131K tokens, Gemma 4 26B A4B 262K.

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 12BBarsGemma 4 26B A4BTies
Coding05
Intelligence02
Agentic work02
Reasoning06
Math03
Knowledge04
General014
All shared boards036
CodingLike-for-likeSix of these boards feed the Vibe Coding Index05
Coding benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
LMArena WebDevNot measured1360Only Gemma 4 26B A4B measured
AA Coding Index5.839.3Gemma 4 26B A4B
LiveCodeBench (Artificial Analysis run)13.7%Not measuredOnly Gemma 3 12B measured
SciCode16.4%40.0%Gemma 4 26B A4B
Terminal-Bench Hard (Artificial Analysis run)0.8%25.0%Gemma 4 26B A4B
WeirdMLNot measured35.2%Only Gemma 4 26B A4B measured
ALE-BenchNot measured927.2Only Gemma 4 26B A4B measured
LMArena Text: coding12811446Gemma 4 26B A4B
LMArena Text: software and IT13131454Gemma 4 26B A4B
LMArena WebDev: ReactNot measured1351Only Gemma 4 26B A4B measured
LMArena WebDev: HTMLNot measured1362Only Gemma 4 26B A4B measured
SciCode0.0%Not measuredOnly Gemma 3 12B measured
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 Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
AA Intelligence Index3.816.7Gemma 4 26B A4B
Epoch Capabilities Index123.5141.9Gemma 4 26B A4B
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
Agentic work benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
AA Agentic Index0.12.9Gemma 4 26B A4B
τ²-Bench Telecom10.8%43.6%Gemma 4 26B A4B
Berkeley Function Calling Leaderboard (BFCL)30.4%Not measuredOnly Gemma 3 12B measured
Neither model has a result on the other 38 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index06
Reasoning benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
GPQA Diamond (Artificial Analysis run)34.9%79.2%Gemma 4 26B A4B
Humanity's Last Exam (Artificial Analysis run)4.2%19.3%Gemma 4 26B A4B
GPQA Diamond (Epoch AI run)39.5%73.2%Gemma 4 26B A4B
LMArena Text: hard prompts13091438Gemma 4 26B A4B
LMArena Text: expert12481447Gemma 4 26B A4B
LMArena Text: hard prompts in English13181448Gemma 4 26B A4B
LMArena Vision: diagramsNot measured1264Only Gemma 4 26B A4B measured
LMArena Vision: homeworkNot measured1293Only Gemma 4 26B A4B measured
Neither model has a result on the other 14 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index03
Math benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
AA Math Index18.3Not measuredOnly Gemma 3 12B measured
AIME 2025 (Artificial Analysis run)18.3%Not measuredOnly Gemma 3 12B measured
OTIS Mock AIME16.7%82.2%Gemma 4 26B A4B
LMArena Text: math13071470Gemma 4 26B A4B
LMArena Text: mathematical work13361475Gemma 4 26B A4B
Neither model has a result on the other 8 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
Knowledge benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
MMLU Pro (Artificial Analysis run)59.5%Not measuredOnly Gemma 3 12B measured
LMArena Text: science13661446Gemma 4 26B A4B
LMArena Text: medicine and healthcare13341428Gemma 4 26B A4B
LMArena Text: legal and government13811457Gemma 4 26B A4B
LMArena Text: business and finance13761429Gemma 4 26B A4B
FACTS Benchmark SuiteNot measured42.7%Only Gemma 4 26B A4B measured
Neither model has a result on the other 12 boards here
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Web design benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index014
General benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
IFBench36.7%72.4%Gemma 4 26B A4B
Long context reasoning (AA-LCR)8.3%65.7%Gemma 4 26B A4B
DTBench48.8%74.9%Gemma 4 26B A4B
LMCA4.5%29.7%Gemma 4 26B A4B
LMArena Text13341434Gemma 4 26B A4B
LMArena Text: instruction following12991420Gemma 4 26B A4B
LMArena Text: creative writing13311400Gemma 4 26B A4B
LMArena VisionNot measured1260Only Gemma 4 26B A4B measured
LMArena Text: English13491444Gemma 4 26B A4B
LMArena Text: non-English13181421Gemma 4 26B A4B
LMArena Text: ChineseNot measured1495Only Gemma 4 26B A4B measured
LMArena Text: German1370Not measuredOnly Gemma 3 12B measured
LMArena Text: FrenchNot measured1460Only Gemma 4 26B A4B measured
LMArena Text: SpanishNot measured1415Only Gemma 4 26B A4B measured
LMArena Text: Russian13351434Gemma 4 26B A4B
LMArena Text: multi-turn13341441Gemma 4 26B A4B
LMArena Text: longer queries13171427Gemma 4 26B A4B
LMArena Text: writing and language13071405Gemma 4 26B A4B
LMArena Text: entertainment and media12971403Gemma 4 26B A4B
LMArena Vision: EnglishNot measured1256Only Gemma 4 26B A4B measured
LMArena Vision: ChineseNot measured1312Only Gemma 4 26B A4B measured
LMArena Vision: reading textNot measured1268Only Gemma 4 26B A4B measured
LMArena Vision: humorNot measured1272Only Gemma 4 26B A4B measured
LMArena Vision: captioningNot measured1297Only Gemma 4 26B A4B measured
LMArena Vision: recognizing thingsNot measured1191Only Gemma 4 26B A4B measured
LMArena Vision: creative writingNot measured1256Only Gemma 4 26B A4B measured
Neither model has a result on the other 20 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Gemma 3 12B and Gemma 4 26B A4B
BenchmarkGemma 3 12BGemma 4 26B A4BLeader
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 12B$0.0002Fits in one request
Gemma 4 26B A4B$0.0004Fits in one request

Gemma 3 12B costs less

Repository review

50K input + 3K output tokens

Gemma 3 12B$0.0042Fits in one request
Gemma 4 26B A4B$0.0093Fits in one request

Gemma 3 12B costs less

Agent loop

200K input + 10K output tokens

Gemma 3 12B$0.017Exceeds the context window
Gemma 4 26B A4B$0.036Fits in one request

Gemma 3 12B 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 12B and Gemma 4 26B A4B
SpecificationGoogle Gemma 3 12BGoogle Gemma 4 26B A4B
MakerGoogleGoogle
WeightsOpen weightsOpen weights
ReleasedMarch 13, 2025April 3, 2026
Context window131K262K
Input $/Mtok$0.075$0.15
Output $/Mtok$0.1625$0.60
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it01
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemma 3 12B better than Gemma 4 26B A4B for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Gemma 3 12B for output tokens 3.7x cheaper ($0.1625 vs $0.60 per million) and work measured by AA Coding Index or SciCode, where it leads.

Which is cheaper, Gemma 3 12B or Gemma 4 26B A4B?

Gemma 3 12B costs $0.075 per million input tokens and $0.1625 per million output tokens; Gemma 4 26B A4B costs $0.15 and $0.60. Cost per solved task is known for neither of them.

Which has the bigger context window, Gemma 3 12B or Gemma 4 26B A4B?

Gemma 4 26B A4B, with 262K tokens against 131K.

How many benchmarks is the Gemma 3 12B vs Gemma 4 26B A4B comparison based on?

36 benchmarks have a result for both models as of October 22, 2026. Gemma 3 12B's coding index rests on 0 independent benchmarks and Gemma 4 26B A4B's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next