LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Gemma 4 26B A4B vs Gemma 4 31B

Updated October 21, 2026. 54 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Gemma 4 26B A4B Replace Gemma 4 31B
Gemma 4 26B A4BGoogle
Coding
22.8
Overall
16.7
Output $/M
$0.255
9
Gemma 4 31BGoogle
Coding
38.3
Overall
14.7
Output $/M
$0.40
45
vs boards won of 54 shared

Decision reading

Pick Gemma 4 31B for coding: it scores 38.3 on the Vibe Coding Index against 22.8 for Gemma 4 26B A4B.

Pick Gemma 4 26B A4B when output tokens cost 1.6x less, or when what matters is ALE-Bench or AA Intelligence Index, where it leads.

Model AGemma 4 26B A4BGoogle22.8/1001 independent benchmarkNo cost per solved taskModel BGemma 4 31BGoogle38.3/1002 independent benchmarksNo cost per solved task
Shared results
54
Gemma 4 26B A4B only
0
Gemma 4 31B only
14
Like-for-like categories
7 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
Gemma 4 26B A4B 22.8 from 1 benchmark, Gemma 4 31B 38.3 from 2 benchmarks.
Widest category
Agentic work, 11.6 points apart on average over 2 shared benchmarks.
Widest single gap
τ²-Bench Telecom: Gemma 4 26B A4B 43.6% against Gemma 4 31B 65.5%, 21.9 points apart on a 0 to 100 scale.
Output price
Gemma 4 26B A4B $0.255 against Gemma 4 31B $0.40 per million output tokens, 1.6x apart.
Context window
Both take 262K tokens.

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 4 26B A4BBarsGemma 4 31BTies
Coding19
Intelligence11
Agentic work02
Reasoning08
Math30
Knowledge13
General322
All shared boards945
CodingLike-for-likeFour of these boards feed the Vibe Coding Index19
Coding benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
LiveBench codingNot measured49.0%Only Gemma 4 31B measured
LMArena WebDev13601367Gemma 4 31B
AA Coding Index39.343.4Gemma 4 31B
SciCode40.0%45.5%Gemma 4 31B
Terminal-Bench Hard (Artificial Analysis run)25.0%36.4%Gemma 4 31B
WeirdML35.2%52.3%Gemma 4 31B
ALE-Bench927.2925.5Gemma 4 26B A4B
LMArena Text: coding14461459Gemma 4 31B
LMArena Text: software and IT14541463Gemma 4 31B
LMArena WebDev: React13511359Gemma 4 31B
LMArena WebDev: HTML13621380Gemma 4 31B
Neither model has a result on the other 27 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index11
Intelligence benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
AA Intelligence Index16.714.7Gemma 4 26B A4B
Epoch Capabilities Index141.9143.0Gemma 4 31B
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
Agentic work benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
AA Agentic Index2.94.2Gemma 4 31B
LiveBench researchNot measured44.4%Only Gemma 4 31B measured
LiveBench browserNot measured4.0%Only Gemma 4 31B measured
LiveBench long agenticNot measured13.3%Only Gemma 4 31B measured
τ²-Bench Telecom43.6%65.5%Gemma 4 31B
Neither model has a result on the other 34 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index08
Reasoning benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
LiveBench reasoningNot measured68.3%Only Gemma 4 31B measured
GPQA Diamond (Artificial Analysis run)79.2%85.7%Gemma 4 31B
Humanity's Last Exam (Artificial Analysis run)19.3%23.6%Gemma 4 31B
GPQA Diamond (Epoch AI run)73.2%75.8%Gemma 4 31B
LMArena Text: hard prompts14381448Gemma 4 31B
LMArena Text: expert14471464Gemma 4 31B
LMArena Text: hard prompts in English14481456Gemma 4 31B
LMArena Vision: diagrams12641283Gemma 4 31B
LMArena Vision: homework12931310Gemma 4 31B
Neither model has a result on the other 12 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index30
Math benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
OTIS Mock AIME82.2%73.3%Gemma 4 26B A4B
LMArena Text: math14701463Gemma 4 26B A4B
LMArena Text: mathematical work14751465Gemma 4 26B A4B
Neither model has a result on the other 10 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index13
Knowledge benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
CaseLaw (Vals AI)Not measured52.6%Only Gemma 4 31B measured
MortgageTax (Vals AI)Not measured61.4%Only Gemma 4 31B measured
SimpleQA VerifiedNot measured10.4%Only Gemma 4 31B measured
LMArena Text: science14461457Gemma 4 31B
LMArena Text: medicine and healthcare14281451Gemma 4 31B
LMArena Text: legal and government14571438Gemma 4 26B A4B
LMArena Text: business and finance14291430Gemma 4 31B
Neither model has a result on the other 10 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
Web design benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
LiveBench designNot measured58.4%Only Gemma 4 31B measured
Neither model has a result on the other 8 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index322
General benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
LiveBench data analysisNot measured53.3%Only Gemma 4 31B measured
SAGE (Vals AI)Not measured55.0%Only Gemma 4 31B measured
IFBench72.4%75.6%Gemma 4 31B
Long context reasoning (AA-LCR)65.7%69.7%Gemma 4 31B
DTBench74.9%82.7%Gemma 4 31B
LMCA29.7%39.3%Gemma 4 31B
Surface Evolver BenchNot measured30.6%Only Gemma 4 31B measured
LMArena Text14341443Gemma 4 31B
LMArena Text: instruction following14201433Gemma 4 31B
LMArena Text: creative writing14001417Gemma 4 31B
LMArena Vision12601276Gemma 4 31B
LMArena Text: English14441451Gemma 4 31B
LMArena Text: non-English14211430Gemma 4 31B
LMArena Text: Chinese14951475Gemma 4 26B A4B
LMArena Text: French14601436Gemma 4 26B A4B
LMArena Text: Spanish14151442Gemma 4 31B
LMArena Text: Russian14341460Gemma 4 31B
LMArena Text: PolishNot measured1433Only Gemma 4 31B measured
LMArena Text: multi-turn14411452Gemma 4 31B
LMArena Text: longer queries14271446Gemma 4 31B
LMArena Text: writing and language14051432Gemma 4 31B
LMArena Text: entertainment and media14031405Gemma 4 31B
LMArena Vision: English12561271Gemma 4 31B
LMArena Vision: Chinese13121326Gemma 4 31B
LMArena Vision: reading text12681282Gemma 4 31B
LMArena Vision: humor12721280Gemma 4 31B
LMArena Vision: captioning12971218Gemma 4 26B A4B
LMArena Vision: recognizing things11911226Gemma 4 31B
LMArena Vision: creative writing12561285Gemma 4 31B
LMArena DocumentNot measured1425Only Gemma 4 31B measured
Neither model has a result on the other 16 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Gemma 4 26B A4B and Gemma 4 31B
BenchmarkGemma 4 26B A4BGemma 4 31BLeader
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 4 26B A4B$0.0002Fits in one request
Gemma 4 31B$0.0004Fits in one request

Gemma 4 26B A4B costs less

Repository review

50K input + 3K output tokens

Gemma 4 26B A4B$0.0046Fits in one request
Gemma 4 31B$0.011Fits in one request

Gemma 4 26B A4B costs less

Agent loop

200K input + 10K output tokens

Gemma 4 26B A4B$0.018Fits in one request
Gemma 4 31B$0.044Fits in one request

Gemma 4 26B A4B 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 4 26B A4B and Gemma 4 31B
SpecificationGoogle Gemma 4 26B A4BGoogle Gemma 4 31B
MakerGoogleGoogle
WeightsOpen weightsOpen weights
ReleasedApril 3, 2026April 2, 2026
Context window262K262K
Input $/Mtok$0.0765$0.20
Output $/Mtok$0.255$0.40
Coding index22.838.3
Independent measurements behind it12
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemma 4 26B A4B better than Gemma 4 31B for coding?

Pick Gemma 4 31B for coding: it scores 38.3 on the Vibe Coding Index against 22.8 for Gemma 4 26B A4B. Pick Gemma 4 26B A4B when output tokens cost 1.6x less, or when what matters is ALE-Bench or AA Intelligence Index, where it leads.

Which is cheaper, Gemma 4 26B A4B or Gemma 4 31B?

Gemma 4 26B A4B costs $0.0765 per million input tokens and $0.255 per million output tokens; Gemma 4 31B costs $0.20 and $0.40. Cost per solved task is known for neither of them.

Which has the bigger context window, Gemma 4 26B A4B or Gemma 4 31B?

Neither: both take 262K tokens.

How many benchmarks is the Gemma 4 26B A4B vs Gemma 4 31B comparison based on?

54 benchmarks have a result for both models as of October 21, 2026. Gemma 4 26B A4B's coding index rests on 1 independent benchmark and Gemma 4 31B's on 2. Where only one model has a result, the page says so instead of naming a winner.

Compare next