Enter opens the highlighted result. Esc closes.

Gemma 4 31B vs Kimi K2.6

Kimi K2.6 leads Gemma 4 31B on coding, 52.1 to 39.0 on the Vibe Coding Index, and Gemma 4 31B costs 10.0x less per output token ($0.40 vs $4.00 per million). Updated October 21, 2026; 58 benchmarks have a result for both.

Replace Gemma 4 31B Replace Kimi K2.6
Gemma 4 31BGoogle
Coding
39.0
Overall
42.1
Output $/M
$0.40
4
Kimi K2.6Moonshot AI
Coding
52.1
Overall
60.5
Output $/M
$4.00
54
vs boards won of 58 shared

Gemma 4 31B or Kimi K2.6: which to pick

Pick Kimi K2.6 for coding: it scores 52.1 on the Vibe Coding Index against 39.0 for Gemma 4 31B.

Pick Gemma 4 31B for output tokens 10.0x cheaper ($0.40 vs $4.00 per million) and work measured by SAGE (Vals AI) or LMCA, where it leads.

Model AGemma 4 31BGoogle39.0/1002 independent benchmarksNo cost per solved taskModel BKimi K2.6Moonshot AI52.1/1003 independent benchmarks$0.35 per solved task
Shared results
58
Gemma 4 31B only
13
Kimi K2.6 only
51
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 4 31B 39.0 from 2 benchmarks, Kimi K2.6 52.1 from 3 benchmarks.
Widest category
Agentic work, 23.4 points apart on average over 2 shared benchmarks.
Widest single gap
τ²-Bench Telecom: Gemma 4 31B 65.5% against Kimi K2.6 95.9%, 30.4 points apart on a 0 to 100 scale.
Output price
Gemma 4 31B $0.40 against Kimi K2.6 $4.00 per million output tokens, 10.0x 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 31BBarsKimi K2.6Ties
Coding010
Intelligence02
Agentic work02
Reasoning08
Math03
Knowledge07
General422
All shared boards454
CodingLike-for-likeSix of these boards feed the Vibe Coding Index010
Coding benchmarks for Gemma 4 31B and Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
LiveBench coding49.0%Not measuredOnly Gemma 4 31B measured
LMArena WebDev13671509Kimi K2.6
AA Coding Index43.461.8Kimi K2.6
LiveCodeBench (Vals AI)Not measured86.8%Only Kimi K2.6 measured
SWE-bench Verified (Vals AI)Not measured76.2%Only Kimi K2.6 measured
Vibe Code Bench (Vals AI)Not measured37.9%Only Kimi K2.6 measured
Code migration (Vals AI)Not measured27.8%Only Kimi K2.6 measured
ProgramBench (Vals AI)Not measured0.0%Only Kimi K2.6 measured
Terminal-Bench 2.1 (Vals AI)Not measured53.6%Only Kimi K2.6 measured
SciCode45.5%51.5%Kimi K2.6
Terminal-Bench Hard (Artificial Analysis run)36.4%43.9%Kimi K2.6
WeirdML52.3%55.9%Kimi K2.6
ALE-Bench925.51092.7Kimi K2.6
LMArena Text: coding14591489Kimi K2.6
LMArena Text: software and IT14631484Kimi K2.6
LMArena WebDev: React13591505Kimi K2.6
LMArena WebDev: HTML13801485Kimi K2.6
LMArena WebDev: from an imageNot measured1521Only Kimi K2.6 measured
SciCode8.9%Not 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 Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
AA Intelligence Index14.727.0Kimi K2.6
Epoch Capabilities Index143.0150.9Kimi K2.6
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
Agentic work benchmarks for Gemma 4 31B and Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
AA Agentic Index4.220.5Kimi K2.6
Design Arena: agents building web appsNot measured1268Only Kimi K2.6 measured
Design Arena: agents building full-stack appsNot measured1155Only Kimi K2.6 measured
Design Arena: agents building mobile appsNot measured1157Only Kimi K2.6 measured
Design Arena: agents building Android appsNot measured1160Only Kimi K2.6 measured
Design Arena: agents building Godot gamesNot measured1120Only Kimi K2.6 measured
Design Arena: agents building gamesNot measured1141Only Kimi K2.6 measured
Design Arena: HTML slidesNot measured1184Only Kimi K2.6 measured
Design Arena: PowerPoint slidesNot measured1128Only Kimi K2.6 measured
Design Arena: spreadsheetsNot measured1276Only Kimi K2.6 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
Finance Agent (Vals AI)Not measured44.9%Only Kimi K2.6 measured
Tax Agent Bench (Vals AI)Not measured12.4%Only Kimi K2.6 measured
Legal Research Bench (Vals AI)Not measured15.9%Only Kimi K2.6 measured
Excel Modeling Benchmark (Vals AI)Not measured57.9%Only Kimi K2.6 measured
τ²-Bench Telecom65.5%95.9%Kimi K2.6
OSWorld 2.0Not measured4.6%Only Kimi K2.6 measured
Vending-Bench 2Not measured$6,205Only Kimi K2.6 measured
Neither model has a result on the other 21 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index08
Reasoning benchmarks for Gemma 4 31B and Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
GPQA DiamondNot measured89.1%Only Kimi K2.6 measured
LiveBench reasoning68.3%Not measuredOnly Gemma 4 31B measured
MMMU Pro (Vals AI)Not measured86.3%Only Kimi K2.6 measured
GPQA Diamond (Artificial Analysis run)85.7%91.1%Kimi K2.6
Humanity's Last Exam (Artificial Analysis run)23.6%37.5%Kimi K2.6
GPQA Diamond (Epoch AI run)75.8%90.8%Kimi K2.6
Chess puzzlesNot measured20.0%Only Kimi K2.6 measured
Mystery game puzzlesNot measured18.0%Only Kimi K2.6 measured
EBR-benchNot measured2.4%Only Kimi K2.6 measured
LMArena Text: hard prompts14481470Kimi K2.6
LMArena Text: expert14641491Kimi K2.6
LMArena Text: hard prompts in English14561471Kimi K2.6
LMArena Vision: diagrams12831298Kimi K2.6
LMArena Vision: homework13101310Kimi K2.6
Kagi LLM Benchmark63.5%Not measuredOnly Gemma 4 31B measured
Neither model has a result on the other 7 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index03
Math benchmarks for Gemma 4 31B and Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
OTIS Mock AIME73.3%96.1%Kimi K2.6
FrontierMath, tiers 1 to 3Not measured57.2%Only Kimi K2.6 measured
FrontierMath, tier 4Not measured25.6%Only Kimi K2.6 measured
ProofBenchNot measured16.0%Only Kimi K2.6 measured
LMArena Text: math14631474Kimi K2.6
LMArena Text: mathematical work14651480Kimi K2.6
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index07
Knowledge benchmarks for Gemma 4 31B and Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
MMLU ProNot measured87.6%Only Kimi K2.6 measured
LegalBench (Vals AI)Not measured84.7%Only Kimi K2.6 measured
CaseLaw (Vals AI)52.6%61.2%Kimi K2.6
CorpFin (Vals AI)Not measured66.7%Only Kimi K2.6 measured
TaxEval (Vals AI)Not measured74.7%Only Kimi K2.6 measured
MortgageTax (Vals AI)61.4%65.8%Kimi K2.6
MedCode (Vals AI)Not measured40.1%Only Kimi K2.6 measured
SimpleQA Verified10.4%34.9%Kimi K2.6
LMArena Text: science14571468Kimi K2.6
LMArena Text: medicine and healthcare14511457Kimi K2.6
LMArena Text: legal and government14381462Kimi K2.6
LMArena Text: business and finance14301445Kimi K2.6
FACTS Benchmark Suite45.3%Not measuredOnly Gemma 4 31B measured
Neither model has a result on the other 5 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 Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
Design Arena: websitesNot measured1269Only Kimi K2.6 measured
Design Arena: UI componentsNot measured1266Only Kimi K2.6 measured
Design Arena: game devNot measured1251Only Kimi K2.6 measured
Design Arena: data vizNot measured1249Only Kimi K2.6 measured
Design Arena: 3DNot measured1268Only Kimi K2.6 measured
Design Arena: all code categoriesNot measured1272Only Kimi K2.6 measured
Design Arena: SVGNot measured1181Only Kimi K2.6 measured
Design Arena: ASCII artNot measured1165Only Kimi K2.6 measured
LiveBench design58.4%Not measuredOnly Gemma 4 31B measured
GeneralLike-for-likePublic boards, shown as published, kept out of the index422
General benchmarks for Gemma 4 31B and Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
LiveBench data analysis53.3%Not measuredOnly Gemma 4 31B measured
Public Benefits Bench (Vals AI)Not measured56.6%Only Kimi K2.6 measured
SAGE (Vals AI)55.0%50.2%Gemma 4 31B
MedScribe (Vals AI)Not measured78.2%Only Kimi K2.6 measured
Vals Multimodal IndexNot measured56.4%Only Kimi K2.6 measured
IFBench75.6%76.0%Kimi K2.6
Long context reasoning (AA-LCR)69.7%81.0%Kimi K2.6
DTBench82.7%90.9%Kimi K2.6
LMCA39.3%37.3%Gemma 4 31B
Surface Evolver Bench30.6%Not measuredOnly Gemma 4 31B measured
Furniture assemblyNot measured21.7%Only Kimi K2.6 measured
ExploitBenchNot measured18.4%Only Kimi K2.6 measured
LMArena Text14431455Kimi K2.6
LMArena Text: instruction following14331451Kimi K2.6
LMArena Text: creative writing14171434Kimi K2.6
LMArena Vision12761282Kimi K2.6
LMArena Text: English14511460Kimi K2.6
LMArena Text: non-English14301446Kimi K2.6
LMArena Text: Chinese14751520Kimi K2.6
LMArena Text: GermanNot measured1451Only Kimi K2.6 measured
LMArena Text: JapaneseNot measured1443Only Kimi K2.6 measured
LMArena Text: KoreanNot measured1427Only Kimi K2.6 measured
LMArena Text: French14361471Kimi K2.6
LMArena Text: Spanish14421464Kimi K2.6
LMArena Text: Russian14601446Gemma 4 31B
LMArena Text: Polish14331438Kimi K2.6
LMArena Text: multi-turn14521453Kimi K2.6
LMArena Text: longer queries14461468Kimi K2.6
LMArena Text: writing and language14321439Kimi K2.6
LMArena Text: entertainment and media14051428Kimi K2.6
LMArena Vision: English12711287Kimi K2.6
LMArena Vision: Chinese13261346Kimi K2.6
LMArena Vision: reading text12821291Kimi K2.6
LMArena Vision: humor12801280Kimi K2.6
LMArena Vision: captioning1218Not measuredOnly Gemma 4 31B measured
LMArena Vision: recognizing things1226Not measuredOnly Gemma 4 31B measured
LMArena Vision: creative writing12851273Gemma 4 31B
LMArena Document14251451Kimi K2.6
Neither model has a result on the other 8 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 Kimi K2.6
BenchmarkGemma 4 31BKimi K2.6Leader
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
Kimi K2.6$0.0029Fits in one request

Gemma 4 31B costs less

Repository review

50K input + 3K output tokens

Gemma 4 31B$0.0082Fits in one request
Kimi K2.6$0.059Fits in one request

Gemma 4 31B costs less

Agent loop

200K input + 10K output tokens

Gemma 4 31B$0.032Fits in one request
Kimi K2.6$0.230Fits 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 Kimi K2.6
SpecificationGoogle Gemma 4 31BMoonshotAI Kimi K2.6
MakerGoogleMoonshot AI
WeightsOpen weightsOpen weights
ReleasedApril 2, 2026April 20, 2026
Context window262K262K
Input $/Mtok$0.14$0.95
Output $/Mtok$0.40$4.00
Coding index39.052.1
Independent measurements behind it23
Cost per solved taskNot computable$0.35
SourceOpenRouter listingOpenRouter listing

Questions

Is Gemma 4 31B better than Kimi K2.6 for coding?

Pick Kimi K2.6 for coding: it scores 52.1 on the Vibe Coding Index against 39.0 for Gemma 4 31B. Pick Gemma 4 31B for output tokens 10.0x cheaper ($0.40 vs $4.00 per million) and work measured by SAGE (Vals AI) or LMCA, where it leads.

Which is cheaper, Gemma 4 31B or Kimi K2.6?

Gemma 4 31B costs $0.14 per million input tokens and $0.40 per million output tokens; Kimi K2.6 costs $0.95 and $4.00. Cost per solved task is known for Kimi K2.6 only.

Which has the bigger context window, Gemma 4 31B or Kimi K2.6?

Neither: both take 262K tokens.

How many benchmarks is the Gemma 4 31B vs Kimi K2.6 comparison based on?

58 benchmarks have a result for both models as of October 21, 2026. Gemma 4 31B's coding index rests on 2 independent benchmarks and Kimi K2.6's on 3. 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-kimi-k2-6" title="Gemma 4 31B vs Kimi K2.6, 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-kimi-k2-6">LLMBenchmark.ai, Gemma 4 31B vs Kimi K2.6</a></p>