Enter opens the highlighted result. Esc closes.

Gemma 4 31B vs GPT-5.6 Luna

GPT-5.6 Luna leads Gemma 4 31B on coding, 58.0 to 39.0 on the Vibe Coding Index, and Gemma 4 31B costs 3.0x less per output token ($0.40 vs $1.20 per million). Updated October 22, 2026; 65 benchmarks have a result for both.

Replace Gemma 4 31B Replace GPT-5.6 Luna
Gemma 4 31BGoogle
Coding
39.0
Overall
42.1
Output $/M
$0.40
26
GPT-5.6 LunaOpenAI
Coding
58.0
Overall
61.6
Output $/M
$1.20
39
vs boards won of 65 shared

Gemma 4 31B or GPT-5.6 Luna: which to pick

Pick GPT-5.6 Luna for coding: it scores 58.0 on the Vibe Coding Index against 39.0 for Gemma 4 31B.

Pick Gemma 4 31B for output tokens 3.0x cheaper ($0.40 vs $1.20 per million) and work measured by SciCode or LMArena Vision: diagrams, where it leads.

Model AGemma 4 31BGoogle39.0/1002 independent benchmarksNo cost per solved taskModel BGPT-5.6 LunaOpenAI58.0/1006 independent benchmarks$0.04 per solved task
Shared results
65
Gemma 4 31B only
6
GPT-5.6 Luna only
54
Like-for-like categories
8 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 4 31B 39.0 from 2 benchmarks, GPT-5.6 Luna 58.0 from 6 benchmarks.
Widest category
Agentic work, 31.4 points apart on average over 4 shared benchmarks.
Widest single gap
LiveBench browser: Gemma 4 31B 4.0% against GPT-5.6 Luna 68.0%, 64.0 points apart on a 0 to 100 scale.
Output price
Gemma 4 31B $0.40 against GPT-5.6 Luna $1.20 per million output tokens, 3.0x apart.
Context window
Gemma 4 31B takes 262K tokens, GPT-5.6 Luna 1.05M.

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-5.6 LunaTies
Coding110
Intelligence02
Agentic work04
Reasoning37
Math12
Knowledge34
Web design01
General189
All shared boards2639
CodingLike-for-likeSix of these boards feed the Vibe Coding Index110
Coding benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
LiveBench coding49.0%73.6%GPT-5.6 Luna
LMArena WebDev13671518GPT-5.6 Luna
AA Coding Index43.471.4GPT-5.6 Luna
SWE-bench Verified (Vals AI)Not measured93.0%Only GPT-5.6 Luna measured
Terminal-Bench 4 (Vals AI)Not measured11.6%Only GPT-5.6 Luna measured
Vibe Code Bench (Vals AI)Not measured77.1%Only GPT-5.6 Luna measured
Code migration (Vals AI)Not measured44.5%Only GPT-5.6 Luna measured
ProgramBench (Vals AI)Not measured0.0%Only GPT-5.6 Luna measured
IOI (Vals AI)Not measured61.8%Only GPT-5.6 Luna measured
Terminal-Bench 2.1 (Vals AI)Not measured79.0%Only GPT-5.6 Luna measured
Vibe Code Bench 1-100 (Vals AI)Not measured22.6%Only GPT-5.6 Luna measured
SkillsBench (Vals AI)Not measured60.5%Only GPT-5.6 Luna measured
CyberBench (Vals AI)Not measured73.6%Only GPT-5.6 Luna measured
SciCode45.5%53.6%GPT-5.6 Luna
Terminal-Bench Hard (Artificial Analysis run)36.4%Not measuredOnly Gemma 4 31B measured
FrontierCodeNot measured39.8%Only GPT-5.6 Luna measured
WeirdML52.3%60.9%GPT-5.6 Luna
ALE-Bench925.51667.4GPT-5.6 Luna
LMArena Text: coding14591466GPT-5.6 Luna
LMArena Text: software and IT14631463GPT-5.6 Luna
LMArena WebDev: React13591508GPT-5.6 Luna
LMArena WebDev: HTML13801520GPT-5.6 Luna
LMArena WebDev: from an imageNot measured1517Only GPT-5.6 Luna measured
SciCode8.9%6.2%Gemma 4 31B
Neither model has a result on the other 15 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
AA Intelligence Index14.737.3GPT-5.6 Luna
Epoch Capabilities Index143.0156.5GPT-5.6 Luna
Agentic workLike-for-likePublic boards, shown as published, kept out of the index04
Agentic work benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
AA Agentic Index4.242.1GPT-5.6 Luna
LiveBench research44.4%51.0%GPT-5.6 Luna
LiveBench browser4.0%68.0%GPT-5.6 Luna
LiveBench long agentic13.3%30.4%GPT-5.6 Luna
SREBench (Vals AI)Not measured1.5%Only GPT-5.6 Luna measured
Finance Agent (Vals AI)Not measured55.0%Only GPT-5.6 Luna measured
Tax Agent Bench (Vals AI)Not measured24.8%Only GPT-5.6 Luna measured
Legal Research Bench (Vals AI)Not measured36.5%Only GPT-5.6 Luna measured
Excel Modeling Benchmark (Vals AI)Not measured67.1%Only GPT-5.6 Luna measured
BioMysteryBench (Vals AI)Not measured61.5%Only GPT-5.6 Luna measured
Terminal-Bench-Science (Vals AI)Not measured0.0%Only GPT-5.6 Luna measured
Vals Science IndexNot measured25.3%Only GPT-5.6 Luna measured
τ²-Bench Telecom65.5%Not measuredOnly Gemma 4 31B measured
APEX-AgentsNot measured43.0%Only GPT-5.6 Luna measured
BALROGNot measured45.6%Only GPT-5.6 Luna measured
Vending-Bench 2Not measured$4,095Only GPT-5.6 Luna measured
Neither model has a result on the other 25 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index37
Reasoning benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
GPQA DiamondNot measured91.7%Only GPT-5.6 Luna measured
LiveBench reasoning68.3%85.6%GPT-5.6 Luna
MysteryMechanism (Vals AI)Not measured14.4%Only GPT-5.6 Luna measured
MMMU Pro (Vals AI)Not measured85.0%Only GPT-5.6 Luna measured
GPQA Diamond (Artificial Analysis run)85.7%91.1%GPT-5.6 Luna
Humanity's Last Exam (Artificial Analysis run)23.6%39.5%GPT-5.6 Luna
GPQA Diamond (Epoch AI run)75.8%91.6%GPT-5.6 Luna
ARC-AGINot measured88.0%Only GPT-5.6 Luna measured
ARC-AGI-2Not measured59.5%Only GPT-5.6 Luna measured
SimpleBenchNot measured46.8%Only GPT-5.6 Luna measured
Chess puzzlesNot measured43.0%Only GPT-5.6 Luna measured
Mystery game puzzlesNot measured21.0%Only GPT-5.6 Luna measured
LMArena Text: hard prompts14481451GPT-5.6 Luna
LMArena Text: expert14641478GPT-5.6 Luna
LMArena Text: hard prompts in English14561460GPT-5.6 Luna
LMArena Vision: diagrams12831273Gemma 4 31B
LMArena Vision: homework13101282Gemma 4 31B
Kagi LLM Benchmark63.5%49.1%Gemma 4 31B
Neither model has a result on the other 4 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index12
Math benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
OTIS Mock AIME73.3%98.3%GPT-5.6 Luna
FrontierMath, tiers 1 to 3Not measured82.1%Only GPT-5.6 Luna measured
FrontierMath, tier 4Not measured61.0%Only GPT-5.6 Luna measured
ProofBenchNot measured60.0%Only GPT-5.6 Luna measured
LMArena Text: math14631459Gemma 4 31B
LMArena Text: mathematical work14651470GPT-5.6 Luna
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index34
Knowledge benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
MMLU ProNot measured86.0%Only GPT-5.6 Luna measured
LegalBench (Vals AI)Not measured84.0%Only GPT-5.6 Luna measured
CaseLaw (Vals AI)52.6%Not measuredOnly Gemma 4 31B measured
CorpFin (Vals AI)Not measured64.2%Only GPT-5.6 Luna measured
TaxEval (Vals AI)Not measured76.2%Only GPT-5.6 Luna measured
MortgageTax (Vals AI)61.4%67.3%GPT-5.6 Luna
MedCode (Vals AI)Not measured42.4%Only GPT-5.6 Luna measured
SimpleQA Verified10.4%41.0%GPT-5.6 Luna
LMArena Text: science14571433Gemma 4 31B
LMArena Text: medicine and healthcare14511399Gemma 4 31B
LMArena Text: legal and government14381446GPT-5.6 Luna
LMArena Text: business and finance14301439GPT-5.6 Luna
FACTS Benchmark Suite45.3%35.0%Gemma 4 31B
Neither model has a result on the other 5 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index01
Web design benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
Design Arena: websitesNot measured1255Only GPT-5.6 Luna measured
Design Arena: UI componentsNot measured1248Only GPT-5.6 Luna measured
Design Arena: game devNot measured1194Only GPT-5.6 Luna measured
Design Arena: data vizNot measured1211Only GPT-5.6 Luna measured
Design Arena: 3DNot measured1121Only GPT-5.6 Luna measured
Design Arena: all code categoriesNot measured1231Only GPT-5.6 Luna measured
Design Arena: SVGNot measured1194Only GPT-5.6 Luna measured
Design Arena: ASCII artNot measured1142Only GPT-5.6 Luna measured
LiveBench design58.4%76.0%GPT-5.6 Luna
GeneralLike-for-likePublic boards, shown as published, kept out of the index189
General benchmarks for Gemma 4 31B and GPT-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
LiveBench data analysis53.3%78.0%GPT-5.6 Luna
Public Benefits Bench (Vals AI)Not measured61.2%Only GPT-5.6 Luna measured
SAGE (Vals AI)55.0%44.2%Gemma 4 31B
MedScribe (Vals AI)Not measured84.4%Only GPT-5.6 Luna measured
Vals Multimodal IndexNot measured69.1%Only GPT-5.6 Luna measured
IFBench75.6%Not measuredOnly Gemma 4 31B measured
Long context reasoning (AA-LCR)69.7%83.7%GPT-5.6 Luna
DTBench82.7%89.1%GPT-5.6 Luna
LMCA39.3%48.5%GPT-5.6 Luna
Surface Evolver Bench30.6%61.9%GPT-5.6 Luna
Furniture assemblyNot measured42.5%Only GPT-5.6 Luna measured
LMArena Text14431431Gemma 4 31B
LMArena Text: instruction following14331437GPT-5.6 Luna
LMArena Text: creative writing14171396Gemma 4 31B
LMArena Vision12761257Gemma 4 31B
LMArena Text: English14511441Gemma 4 31B
LMArena Text: non-English14301417Gemma 4 31B
LMArena Text: Chinese14751470Gemma 4 31B
LMArena Text: GermanNot measured1455Only GPT-5.6 Luna measured
LMArena Text: JapaneseNot measured1415Only GPT-5.6 Luna measured
LMArena Text: KoreanNot measured1415Only GPT-5.6 Luna measured
LMArena Text: French14361458GPT-5.6 Luna
LMArena Text: Spanish14421446GPT-5.6 Luna
LMArena Text: Russian14601427Gemma 4 31B
LMArena Text: Polish14331405Gemma 4 31B
LMArena Text: multi-turn14521434Gemma 4 31B
LMArena Text: longer queries14461436Gemma 4 31B
LMArena Text: writing and language14321419Gemma 4 31B
LMArena Text: entertainment and media14051402Gemma 4 31B
LMArena Vision: English12711263Gemma 4 31B
LMArena Vision: Chinese13261294Gemma 4 31B
LMArena Vision: reading text12821264Gemma 4 31B
LMArena Vision: humor12801190Gemma 4 31B
LMArena Vision: captioning1218Not measuredOnly Gemma 4 31B measured
LMArena Vision: recognizing things1226Not measuredOnly Gemma 4 31B measured
LMArena Vision: creative writing12851257Gemma 4 31B
LMArena Document14251457GPT-5.6 Luna
Neither model has a result on the other 9 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-5.6 Luna
BenchmarkGemma 4 31BGPT-5.6 LunaLeader
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$0.0003Fits in one request
GPT-5.6 Luna$0.0008Fits in one request

Gemma 4 31B costs less

Repository review

50K input + 3K output tokens

Gemma 4 31B$0.0082Fits in one request
GPT-5.6 Luna$0.014Fits in one request

Gemma 4 31B costs less

Agent loop

200K input + 10K output tokens

Gemma 4 31B$0.032Fits in one request
GPT-5.6 Luna$0.052Fits 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-5.6 Luna
SpecificationGoogle Gemma 4 31BOpenAI GPT-5.6 Luna
MakerGoogleOpenAI
WeightsOpen weightsNo public weights listed
ReleasedApril 2, 2026July 9, 2026
Context window262K1.05M
Input $/Mtok$0.14$0.20
Output $/Mtok$0.40$1.20
Coding index39.058.0
Independent measurements behind it26
Cost per solved taskNot computable$0.04
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemma 4 31B better than GPT-5.6 Luna for coding?

Pick GPT-5.6 Luna for coding: it scores 58.0 on the Vibe Coding Index against 39.0 for Gemma 4 31B. Pick Gemma 4 31B for output tokens 3.0x cheaper ($0.40 vs $1.20 per million) and work measured by SciCode or LMArena Vision: diagrams, where it leads.

Which is cheaper, Gemma 4 31B or GPT-5.6 Luna?

Gemma 4 31B costs $0.14 per million input tokens and $0.40 per million output tokens; GPT-5.6 Luna costs $0.20 and $1.20. Cost per solved task is known for GPT-5.6 Luna only.

Which has the bigger context window, Gemma 4 31B or GPT-5.6 Luna?

GPT-5.6 Luna, with 1.05M tokens against 262K.

How many benchmarks is the Gemma 4 31B vs GPT-5.6 Luna comparison based on?

65 benchmarks have a result for both models as of October 22, 2026. Gemma 4 31B's coding index rests on 2 independent benchmarks and GPT-5.6 Luna's on 6. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/gemma-4-31b-it-vs-gpt-5-6-luna" title="Gemma 4 31B vs GPT-5.6 Luna, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/gemma-4-31b-it-vs-gpt-5-6-luna">LLMBenchmark.ai, Gemma 4 31B vs GPT-5.6 Luna</a></p>