Gemma 4 31B vs Kimi K2 Thinking
Updated October 21, 2026. 11 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick Kimi K2 Thinking for coding: it scores 63.4 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 6.3x less, or when what matters is Terminal-Bench Hard (Artificial Analysis run) or WeirdML, where it leads.
- Shared results
- 11
- Gemma 4 31B only
- 57
- Kimi K2 Thinking only
- 19
- Like-for-like categories
- 6 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.
Coding
Vibe Coding Index, four public coding benchmarks
Kimi K2 Thinking
Kimi K2 Thinking scores 63.4 against 38.3.
2 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
Kimi K2 Thinking
Kimi K2 Thinking leads, 22.0 against 14.7.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Gemma 4 31B has a AA Agentic Index result, 4.2.
Directional only
Reasoning
GPQA Diamond
No call
Only Kimi K2 Thinking has a GPQA Diamond result, 78.5%.
Directional only
Math
AIME
No call
Only Kimi K2 Thinking has a AIME result, 85.4%.
Directional only
Knowledge
MMLU Pro
No call
Only Kimi K2 Thinking has a MMLU Pro result, 81.1%.
Directional only
Web design
Design Arena: websites
No call
Only Kimi K2 Thinking has a Design Arena: websites result, 1113.
Directional only
Long documents
Listed context window
No call
Both take 262K tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Gemma 4 31B
Gemma 4 31B $0.0004, Kimi K2 Thinking $0.0019 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Kimi K2 Thinking has the number, $0.61.
Directional only
At a glance
- Coding index
- Gemma 4 31B 38.3 from 2 benchmarks, Kimi K2 Thinking 63.4 from 1 benchmark.
- Widest category
- Agentic work, 27.5 points apart on average over 1 shared benchmark.
- Widest single gap
- τ²-Bench Telecom: Gemma 4 31B 65.5% against Kimi K2 Thinking 93.0%, 27.5 points apart on a 0 to 100 scale.
- Output price
- Gemma 4 31B $0.40 against Kimi K2 Thinking $2.50 per million output tokens, 6.3x 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.
| Category | Gemma 4 31B | Bars | Kimi K2 Thinking | Ties |
|---|---|---|---|---|
| Coding | 3 | 0 | ||
| Intelligence | 0 | 2 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 1 | 1 | ||
| Knowledge | 0 | 1 | ||
| General | 1 | 1 | ||
| All shared boards | 5 | 6 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index30
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| SWE-bench Verified | Not measured | 63.4% | Only Kimi K2 Thinking measured |
| LiveBench coding | 49.0% | Not measured | Only Gemma 4 31B measured |
| LMArena WebDev | 1367 | Not measured | Only Gemma 4 31B measured |
| AA Coding Index | 43.4 | Not measured | Only Gemma 4 31B measured |
| LiveCodeBench (Vals AI) | Not measured | 63.1% | Only Kimi K2 Thinking measured |
| SWE-bench Verified (Vals AI) | Not measured | 60.2% | Only Kimi K2 Thinking measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 85.3% | Only Kimi K2 Thinking measured |
| SciCode | 45.5% | Not measured | Only Gemma 4 31B measured |
| Terminal-Bench Hard (Artificial Analysis run) | 36.4% | 31.1% | Gemma 4 31B |
| Terminal-Bench | Not measured | 35.7% | Only Kimi K2 Thinking measured |
| WeirdML | 52.3% | 42.8% | Gemma 4 31B |
| ALE-Bench | 925.5 | 597.5 | Gemma 4 31B |
| LMArena Text: coding | 1459 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: software and IT | 1463 | Not measured | Only Gemma 4 31B measured |
| LMArena WebDev: React | 1359 | Not measured | Only Gemma 4 31B measured |
| LMArena WebDev: HTML | 1380 | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 22 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| AA Intelligence Index | 14.7 | 22.0 | Kimi K2 Thinking |
| Epoch Capabilities Index | 143.0 | 145.8 | Kimi K2 Thinking |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| AA Agentic Index | 4.2 | Not measured | Only Gemma 4 31B measured |
| LiveBench research | 44.4% | Not measured | Only Gemma 4 31B measured |
| LiveBench browser | 4.0% | Not measured | Only Gemma 4 31B measured |
| LiveBench long agentic | 13.3% | Not measured | Only Gemma 4 31B measured |
| τ²-Bench Telecom | 65.5% | 93.0% | Kimi K2 Thinking |
| METR time horizons | Not measured | 59.2% | Only Kimi K2 Thinking measured |
| Neither model has a result on the other 33 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 78.5% | Only Kimi K2 Thinking measured |
| LiveBench reasoning | 68.3% | Not measured | Only Gemma 4 31B measured |
| GPQA Diamond (Artificial Analysis run) | 85.7% | 83.8% | Gemma 4 31B |
| Humanity's Last Exam (Artificial Analysis run) | 23.6% | 23.8% | Kimi K2 Thinking |
| GPQA Diamond (Epoch AI run) | 75.8% | Not measured | Only Gemma 4 31B measured |
| LMArena Text: hard prompts | 1448 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: expert | 1464 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: hard prompts in English | 1456 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: diagrams | 1283 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: homework | 1310 | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 11 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| AIME | Not measured | 85.4% | Only Kimi K2 Thinking measured |
| MGSM (Vals AI) | Not measured | 90.2% | Only Kimi K2 Thinking measured |
| AA Math Index | Not measured | 94.7 | Only Kimi K2 Thinking measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 94.7% | Only Kimi K2 Thinking measured |
| OTIS Mock AIME | 73.3% | Not measured | Only Gemma 4 31B measured |
| LMArena Text: math | 1463 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: mathematical work | 1465 | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 81.1% | Only Kimi K2 Thinking measured |
| LegalBench (Vals AI) | Not measured | 80.2% | Only Kimi K2 Thinking measured |
| CaseLaw (Vals AI) | 52.6% | 65.7% | Kimi K2 Thinking |
| CorpFin (Vals AI) | Not measured | 60.6% | Only Kimi K2 Thinking measured |
| TaxEval (Vals AI) | Not measured | 71.7% | Only Kimi K2 Thinking measured |
| MortgageTax (Vals AI) | 61.4% | Not measured | Only Gemma 4 31B measured |
| MedQA (Vals AI) | Not measured | 92.6% | Only Kimi K2 Thinking measured |
| MMLU Pro (Artificial Analysis run) | Not measured | 84.8% | Only Kimi K2 Thinking measured |
| SimpleQA Verified | 10.4% | Not measured | Only Gemma 4 31B measured |
| LMArena Text: science | 1457 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: medicine and healthcare | 1451 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: legal and government | 1438 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: business and finance | 1430 | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 4 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1113 | Only Kimi K2 Thinking measured |
| LiveBench design | 58.4% | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 7 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| LiveBench data analysis | 53.3% | Not measured | Only Gemma 4 31B measured |
| SAGE (Vals AI) | 55.0% | Not measured | Only Gemma 4 31B measured |
| IFBench | 75.6% | 68.1% | Gemma 4 31B |
| Long context reasoning (AA-LCR) | 69.7% | 72.0% | Kimi K2 Thinking |
| DTBench | 82.7% | Not measured | Only Gemma 4 31B measured |
| LMCA | 39.3% | Not measured | Only Gemma 4 31B measured |
| CL-bench | Not measured | 17.6% | Only Kimi K2 Thinking measured |
| Surface Evolver Bench | 30.6% | Not measured | Only Gemma 4 31B measured |
| LMArena Text | 1443 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: instruction following | 1433 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: creative writing | 1417 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision | 1276 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: English | 1451 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: non-English | 1430 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: Chinese | 1475 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: French | 1436 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: Spanish | 1442 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: Russian | 1460 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: Polish | 1433 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: multi-turn | 1452 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: longer queries | 1446 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: writing and language | 1432 | Not measured | Only Gemma 4 31B measured |
| LMArena Text: entertainment and media | 1405 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: English | 1271 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: Chinese | 1326 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: reading text | 1282 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: humor | 1280 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: captioning | 1218 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: recognizing things | 1226 | Not measured | Only Gemma 4 31B measured |
| LMArena Vision: creative writing | 1285 | Not measured | Only Gemma 4 31B measured |
| LMArena Document | 1425 | Not measured | Only Gemma 4 31B measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemma 4 31B | Kimi K2 Thinking | Leader |
|---|---|---|---|
| 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 costs less
Repository review
50K input + 3K output tokens
Gemma 4 31B costs less
Agent loop
200K input + 10K output tokens
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.
| Specification | Google Gemma 4 31B | MoonshotAI Kimi K2 Thinking |
|---|---|---|
| Maker | Moonshot AI | |
| Weights | Open weights | Open weights |
| Released | April 2, 2026 | November 6, 2025 |
| Context window | 262K | 262K |
| Input $/Mtok | $0.20 | $0.60 |
| Output $/Mtok | $0.40 | $2.50 |
| Coding index | 38.3 | 63.4 |
| Independent measurements behind it | 2 | 1 |
| Cost per solved task | Not computable | $0.61 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemma 4 31B better than Kimi K2 Thinking for coding?
Pick Kimi K2 Thinking for coding: it scores 63.4 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 6.3x less, or when what matters is Terminal-Bench Hard (Artificial Analysis run) or WeirdML, where it leads.
Which is cheaper, Gemma 4 31B or Kimi K2 Thinking?
Gemma 4 31B costs $0.20 per million input tokens and $0.40 per million output tokens; Kimi K2 Thinking costs $0.60 and $2.50. Cost per solved task is known for Kimi K2 Thinking only.
Which has the bigger context window, Gemma 4 31B or Kimi K2 Thinking?
Neither: both take 262K tokens.
How many benchmarks is the Gemma 4 31B vs Kimi K2 Thinking comparison based on?
11 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 Thinking's on 1. Where only one model has a result, the page says so instead of naming a winner.