LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Gemma 4 31B vs GPT-5 Mini

Updated October 21, 2026. 53 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Gemma 4 31B Replace GPT-5 Mini
Gemma 4 31BGoogle
Coding
38.3
Overall
14.7
Output $/M
$0.40
43
GPT-5 MiniOpenAI
Coding
59.8
Overall
20.6
Output $/M
$2.00
10
vs boards won of 53 shared

Decision reading

Pick GPT-5 Mini for coding: it scores 59.8 on the Vibe Coding Index against 38.3 for Gemma 4 31B, though on 1 benchmark against 2.

Pick Gemma 4 31B when output tokens cost 5.0x less, or when what matters is SciCode or Terminal-Bench Hard (Artificial Analysis run), where it leads.

Model AGemma 4 31BGoogle38.3/1002 independent benchmarksNo cost per solved taskModel BGPT-5 MiniOpenAI59.8/1001 independent benchmark$0.08 per solved task
Shared results
53
Gemma 4 31B only
15
GPT-5 Mini only
45
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 31B 38.3 from 2 benchmarks, GPT-5 Mini 59.8 from 1 benchmark.
Widest category
Math, 13.9 points apart on average over 3 shared benchmarks.
Widest single gap
LMArena Vision: Chinese: Gemma 4 31B 1326 against GPT-5 Mini 1214, 18.7 points apart on a 0 to 100 scale.
Output price
Gemma 4 31B $0.40 against GPT-5 Mini $2.00 per million output tokens, 5.0x apart.
Context window
Gemma 4 31B takes 262K tokens, GPT-5 Mini 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 4 31BBarsGPT-5 MiniTies
Coding41
Intelligence02
Agentic work01
Reasoning80
Math21
Knowledge43
General252
All shared boards4310
CodingLike-for-likeFour of these boards feed the Vibe Coding Index41
Coding benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
SWE-bench VerifiedNot measured59.8%Only GPT-5 Mini measured
SWE-bench MultilingualNot measured39.7%Only GPT-5 Mini measured
LiveBench coding49.0%Not measuredOnly Gemma 4 31B measured
LMArena WebDev1367Not measuredOnly Gemma 4 31B measured
AA Coding Index43.4Not measuredOnly Gemma 4 31B measured
LiveCodeBench (Vals AI)Not measured86.6%Only GPT-5 Mini measured
SWE-bench Verified (Vals AI)Not measured60.8%Only GPT-5 Mini measured
Vibe Code Bench (Vals AI)Not measured14.2%Only GPT-5 Mini measured
LiveCodeBench (Artificial Analysis run)Not measured83.8%Only GPT-5 Mini measured
SciCode45.5%39.0%Gemma 4 31B
Terminal-Bench Hard (Artificial Analysis run)36.4%33.3%Gemma 4 31B
Terminal-BenchNot measured34.8%Only GPT-5 Mini measured
WeirdML52.3%52.7%GPT-5 Mini
ALE-Bench925.5Not measuredOnly Gemma 4 31B measured
LMArena Text: coding14591406Gemma 4 31B
LMArena Text: software and IT14631394Gemma 4 31B
LMArena WebDev: React1359Not measuredOnly Gemma 4 31B measured
LMArena WebDev: HTML1380Not measuredOnly Gemma 4 31B measured
Neither model has a result on the other 20 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
AA Intelligence Index14.720.6GPT-5 Mini
Epoch Capabilities Index143.0145.3GPT-5 Mini
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
Agentic work benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
AA Agentic Index4.2Not measuredOnly Gemma 4 31B measured
LiveBench research44.4%Not measuredOnly Gemma 4 31B measured
LiveBench browser4.0%Not measuredOnly Gemma 4 31B measured
LiveBench long agentic13.3%Not measuredOnly Gemma 4 31B measured
τ²-Bench Telecom65.5%71.1%GPT-5 Mini
Vending-Bench 2Not measured-$31Only GPT-5 Mini measured
Neither model has a result on the other 33 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index80
Reasoning benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
GPQA DiamondNot measured80.3%Only GPT-5 Mini measured
LiveBench reasoning68.3%Not measuredOnly Gemma 4 31B measured
MMMU Pro (Vals AI)Not measured78.9%Only GPT-5 Mini measured
GPQA Diamond (Artificial Analysis run)85.7%82.8%Gemma 4 31B
Humanity's Last Exam (Artificial Analysis run)23.6%21.5%Gemma 4 31B
GPQA Diamond (Epoch AI run)75.8%75.0%Gemma 4 31B
ARC-AGINot measured54.3%Only GPT-5 Mini measured
ARC-AGI-2Not measured4.4%Only GPT-5 Mini measured
Chess puzzlesNot measured24.0%Only GPT-5 Mini measured
Mystery game puzzlesNot measured10.0%Only GPT-5 Mini measured
VPCTNot measured40.2%Only GPT-5 Mini measured
LMArena Text: hard prompts14481380Gemma 4 31B
LMArena Text: expert14641379Gemma 4 31B
LMArena Text: hard prompts in English14561383Gemma 4 31B
LMArena Vision: diagrams12831202Gemma 4 31B
LMArena Vision: homework13101210Gemma 4 31B
Neither model has a result on the other 5 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index21
Math benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
AIMENot measured91.5%Only GPT-5 Mini measured
MATH 500 (Vals AI)Not measured94.8%Only GPT-5 Mini measured
MGSM (Vals AI)Not measured92.6%Only GPT-5 Mini measured
AA Math IndexNot measured90.7Only GPT-5 Mini measured
AIME 2025 (Artificial Analysis run)Not measured90.7%Only GPT-5 Mini measured
OTIS Mock AIME73.3%86.7%GPT-5 Mini
FrontierMath, tiers 1 to 3Not measured46.7%Only GPT-5 Mini measured
FrontierMath, tier 4Not measured12.2%Only GPT-5 Mini measured
ProofBenchNot measured9.0%Only GPT-5 Mini measured
MATH level 5Not measured97.8%Only GPT-5 Mini measured
LMArena Text: math14631378Gemma 4 31B
LMArena Text: mathematical work14651380Gemma 4 31B
Neither model has a result on the other 1 board here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index43
Knowledge benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
MMLU ProNot measured82.2%Only GPT-5 Mini measured
LegalBench (Vals AI)Not measured81.8%Only GPT-5 Mini measured
CaseLaw (Vals AI)52.6%68.5%GPT-5 Mini
CorpFin (Vals AI)Not measured60.2%Only GPT-5 Mini measured
TaxEval (Vals AI)Not measured75.2%Only GPT-5 Mini measured
MortgageTax (Vals AI)61.4%66.9%GPT-5 Mini
MedQA (Vals AI)Not measured96.1%Only GPT-5 Mini measured
MedCode (Vals AI)Not measured43.0%Only GPT-5 Mini measured
MMLU Pro (Artificial Analysis run)Not measured83.7%Only GPT-5 Mini measured
SimpleQA Verified10.4%21.6%GPT-5 Mini
LMArena Text: science14571371Gemma 4 31B
LMArena Text: medicine and healthcare14511380Gemma 4 31B
LMArena Text: legal and government14381375Gemma 4 31B
LMArena Text: business and finance14301358Gemma 4 31B
Neither model has a result on the other 3 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
Web design benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
Design Arena: websitesNot measured1125Only GPT-5 Mini measured
Design Arena: UI componentsNot measured1106Only GPT-5 Mini measured
Design Arena: game devNot measured1130Only GPT-5 Mini measured
Design Arena: data vizNot measured1131Only GPT-5 Mini measured
Design Arena: 3DNot measured1037Only GPT-5 Mini measured
Design Arena: all code categoriesNot measured1115Only GPT-5 Mini measured
Design Arena: SVGNot measured1091Only GPT-5 Mini measured
Design Arena: ASCII artNot measured1124Only GPT-5 Mini measured
LiveBench design58.4%Not measuredOnly Gemma 4 31B measured
GeneralLike-for-likePublic boards, shown as published, kept out of the index252
General benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
LiveBench data analysis53.3%Not measuredOnly Gemma 4 31B measured
SAGE (Vals AI)55.0%43.0%Gemma 4 31B
MedScribe (Vals AI)Not measured80.6%Only GPT-5 Mini measured
IFBench75.6%75.4%Gemma 4 31B
Long context reasoning (AA-LCR)69.7%72.3%GPT-5 Mini
DTBench82.7%80.5%Gemma 4 31B
LMCA39.3%34.2%Gemma 4 31B
Fiction.LiveBenchNot measured69.4%Only GPT-5 Mini measured
Creative writing (Lech Mazur)Not measured83.1%Only GPT-5 Mini measured
Surface Evolver Bench30.6%Not measuredOnly Gemma 4 31B measured
LMArena Text14431373Gemma 4 31B
LMArena Text: instruction following14331357Gemma 4 31B
LMArena Text: creative writing14171325Gemma 4 31B
LMArena Vision12761202Gemma 4 31B
LMArena Text: English14511379Gemma 4 31B
LMArena Text: non-English14301363Gemma 4 31B
LMArena Text: Chinese14751386Gemma 4 31B
LMArena Text: GermanNot measured1366Only GPT-5 Mini measured
LMArena Text: JapaneseNot measured1341Only GPT-5 Mini measured
LMArena Text: KoreanNot measured1308Only GPT-5 Mini measured
LMArena Text: French14361386Gemma 4 31B
LMArena Text: Spanish14421354Gemma 4 31B
LMArena Text: Russian14601362Gemma 4 31B
LMArena Text: Polish14331378Gemma 4 31B
LMArena Text: multi-turn14521363Gemma 4 31B
LMArena Text: longer queries14461355Gemma 4 31B
LMArena Text: writing and language14321344Gemma 4 31B
LMArena Text: entertainment and media14051340Gemma 4 31B
LMArena Vision: English12711216Gemma 4 31B
LMArena Vision: Chinese13261214Gemma 4 31B
LMArena Vision: reading text12821199Gemma 4 31B
LMArena Vision: humor12801205Gemma 4 31B
LMArena Vision: captioning12181217Gemma 4 31B
LMArena Vision: recognizing things12261226GPT-5 Mini
LMArena Vision: creative writing12851181Gemma 4 31B
LMArena Document1425Not measuredOnly Gemma 4 31B measured
Neither model has a result on the other 10 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Gemma 4 31B and GPT-5 Mini
BenchmarkGemma 4 31BGPT-5 MiniLeader
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 31B$0.0004Fits in one request
GPT-5 Mini$0.0013Fits in one request

Gemma 4 31B costs less

Repository review

50K input + 3K output tokens

Gemma 4 31B$0.011Fits in one request
GPT-5 Mini$0.018Fits in one request

Gemma 4 31B costs less

Agent loop

200K input + 10K output tokens

Gemma 4 31B$0.044Fits in one request
GPT-5 Mini$0.070Fits in one request

Gemma 4 31B 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 31B and GPT-5 Mini
SpecificationGoogle Gemma 4 31BOpenAI GPT-5 Mini
MakerGoogleOpenAI
WeightsOpen weightsNo public weights listed
ReleasedApril 2, 2026August 7, 2025
Context window262K400K
Input $/Mtok$0.20$0.25
Output $/Mtok$0.40$2.00
Coding index38.359.8
Independent measurements behind it21
Cost per solved taskNot computable$0.08
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemma 4 31B better than GPT-5 Mini for coding?

Pick GPT-5 Mini for coding: it scores 59.8 on the Vibe Coding Index against 38.3 for Gemma 4 31B, though on 1 benchmark against 2. Pick Gemma 4 31B when output tokens cost 5.0x less, or when what matters is SciCode or Terminal-Bench Hard (Artificial Analysis run), where it leads.

Which is cheaper, Gemma 4 31B or GPT-5 Mini?

Gemma 4 31B costs $0.20 per million input tokens and $0.40 per million output tokens; GPT-5 Mini costs $0.25 and $2.00. Cost per solved task is known for GPT-5 Mini only.

Which has the bigger context window, Gemma 4 31B or GPT-5 Mini?

GPT-5 Mini, with 400K tokens against 262K.

How many benchmarks is the Gemma 4 31B vs GPT-5 Mini comparison based on?

53 benchmarks have a result for both models as of October 21, 2026. Gemma 4 31B's coding index rests on 2 independent benchmarks and GPT-5 Mini's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next