Gemma 3 12B vs Mistral Large 4
Mistral Large 4 leads Gemma 3 12B on the Artificial Analysis Intelligence Index, 38.4 to 3.8, and Gemma 3 12B costs 25.7x less per output token ($0.1625 vs $4.18 per million). Updated October 21, 2026; 30 benchmarks have a result for both.
Gemma 3 12B or Mistral Large 4: which to pick
Only Mistral Large 4 has a coding score, 56.4 on the Vibe Coding Index; Gemma 3 12B has none, so the two cannot be ranked on coding.
Pick Gemma 3 12B for output tokens 25.7x cheaper ($0.1625 vs $4.18 per million).
- Shared results
- 30
- Gemma 3 12B only
- 13
- Mistral Large 4 only
- 51
- Like-for-like categories
- 6 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.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Only Mistral Large 4 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Mistral Large 4
Mistral Large 4 leads, 38.4 against 3.8.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Gemma 3 12B has a AA Agentic Index result, 0.1.
Directional only
Web design
Design Arena: websites
No call
Only Mistral Large 4 has a Design Arena: websites result, 1258.
Directional only
Long documents
Listed context window
Mistral Large 4
Mistral Large 4 takes 1.05M tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Gemma 3 12B
Gemma 3 12B $0.0002, Mistral Large 4 $0.0034 per turn.
List prices
At a glance
- Coding index
- Gemma 3 12B not benchmarked, Mistral Large 4 56.4 from 5 benchmarks.
- Widest category
- Math, 43.2 points apart on average over 3 shared benchmarks.
- Widest single gap
- OTIS Mock AIME: Gemma 3 12B 16.7% against Mistral Large 4 98.3%, 81.7 points apart on a 0 to 100 scale.
- Output price
- Gemma 3 12B $0.1625 against Mistral Large 4 $4.18 per million output tokens, 25.7x apart.
- Context window
- Gemma 3 12B takes 131K tokens, Mistral Large 4 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.
| Category | Gemma 3 12B | Bars | Mistral Large 4 | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 2 | ||
| Reasoning | 0 | 5 | ||
| Math | 0 | 3 | ||
| Knowledge | 0 | 4 | ||
| General | 0 | 13 | ||
| All shared boards | 0 | 30 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index03
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 60.0% | Only Mistral Large 4 measured |
| LMArena WebDev | Not measured | 1543 | Only Mistral Large 4 measured |
| AA Coding Index | 5.8 | Not measured | Only Gemma 3 12B measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 27.8% | Only Mistral Large 4 measured |
| Vibe Code Bench (Vals AI) | Not measured | 78.4% | Only Mistral Large 4 measured |
| Code migration (Vals AI) | Not measured | 30.6% | Only Mistral Large 4 measured |
| IOI (Vals AI) | Not measured | 45.3% | Only Mistral Large 4 measured |
| SWE-bench Multimodal (Vals AI) | Not measured | 33.1% | Only Mistral Large 4 measured |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 14.3% | Only Mistral Large 4 measured |
| LiveCodeBench (Artificial Analysis run) | 13.7% | Not measured | Only Gemma 3 12B measured |
| SciCode | 16.4% | 54.2% | Mistral Large 4 |
| Terminal-Bench Hard (Artificial Analysis run) | 0.8% | Not measured | Only Gemma 3 12B measured |
| FrontierCode | Not measured | 39.4% | Only Mistral Large 4 measured |
| ALE-Bench | Not measured | 1213.2 | Only Mistral Large 4 measured |
| LMArena Text: coding | 1281 | 1478 | Mistral Large 4 |
| LMArena Text: software and IT | 1313 | 1467 | Mistral Large 4 |
| LMArena WebDev: React | Not measured | 1559 | Only Mistral Large 4 measured |
| LMArena WebDev: HTML | Not measured | 1487 | Only Mistral Large 4 measured |
| SciCode | 0.0% | Not measured | Only Gemma 3 12B measured |
| Neither model has a result on the other 20 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| AA Intelligence Index | 3.8 | 38.4 | Mistral Large 4 |
| Epoch Capabilities Index | 123.5 | 150.0 | Mistral Large 4 |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index12 boards, none shared
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| AA Agentic Index | 0.1 | Not measured | Only Gemma 3 12B measured |
| LiveBench research | Not measured | 28.1% | Only Mistral Large 4 measured |
| LiveBench browser | Not measured | 8.0% | Only Mistral Large 4 measured |
| LiveBench long agentic | Not measured | 11.8% | Only Mistral Large 4 measured |
| Finance Agent (Vals AI) | Not measured | 54.7% | Only Mistral Large 4 measured |
| Tax Agent Bench (Vals AI) | Not measured | 30.4% | Only Mistral Large 4 measured |
| Legal Research Bench (Vals AI) | Not measured | 31.7% | Only Mistral Large 4 measured |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 17.5% | Only Mistral Large 4 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 56.2% | Only Mistral Large 4 measured |
| BioMysteryBench (Vals AI) | Not measured | 67.0% | Only Mistral Large 4 measured |
| τ²-Bench Telecom | 10.8% | Not measured | Only Gemma 3 12B measured |
| Berkeley Function Calling Leaderboard (BFCL) | 30.4% | Not measured | Only Gemma 3 12B measured |
| Neither model has a result on the other 29 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| LiveBench reasoning | Not measured | 83.9% | Only Mistral Large 4 measured |
| MysteryMechanism (Vals AI) | Not measured | 12.2% | Only Mistral Large 4 measured |
| GPQA Diamond (Artificial Analysis run) | 34.9% | Not measured | Only Gemma 3 12B measured |
| Humanity's Last Exam (Artificial Analysis run) | 4.2% | 35.0% | Mistral Large 4 |
| GPQA Diamond (Epoch AI run) | 39.5% | 88.1% | Mistral Large 4 |
| Chess puzzles | Not measured | 33.0% | Only Mistral Large 4 measured |
| Mystery game puzzles | Not measured | 17.0% | Only Mistral Large 4 measured |
| LMArena Text: hard prompts | 1309 | 1444 | Mistral Large 4 |
| LMArena Text: expert | 1248 | 1444 | Mistral Large 4 |
| LMArena Text: hard prompts in English | 1318 | 1442 | Mistral Large 4 |
| LMArena Vision: diagrams | Not measured | 1288 | Only Mistral Large 4 measured |
| Neither model has a result on the other 11 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index03
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| AA Math Index | 18.3 | Not measured | Only Gemma 3 12B measured |
| AIME 2025 (Artificial Analysis run) | 18.3% | Not measured | Only Gemma 3 12B measured |
| OTIS Mock AIME | 16.7% | 98.3% | Mistral Large 4 |
| FrontierMath, tiers 1 to 3 | Not measured | 58.6% | Only Mistral Large 4 measured |
| FrontierMath, tier 4 | Not measured | 26.8% | Only Mistral Large 4 measured |
| ProofBench | Not measured | 10.0% | Only Mistral Large 4 measured |
| LMArena Text: math | 1307 | 1469 | Mistral Large 4 |
| LMArena Text: mathematical work | 1336 | 1462 | Mistral Large 4 |
| Neither model has a result on the other 5 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| MedCode (Vals AI) | Not measured | 40.7% | Only Mistral Large 4 measured |
| MMLU Pro (Artificial Analysis run) | 59.5% | Not measured | Only Gemma 3 12B measured |
| SimpleQA Verified | Not measured | 20.0% | Only Mistral Large 4 measured |
| LMArena Text: science | 1366 | 1448 | Mistral Large 4 |
| LMArena Text: medicine and healthcare | 1334 | 1433 | Mistral Large 4 |
| LMArena Text: legal and government | 1381 | 1425 | Mistral Large 4 |
| LMArena Text: business and finance | 1376 | 1439 | Mistral Large 4 |
| Neither model has a result on the other 11 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1258 | Only Mistral Large 4 measured |
| Design Arena: UI components | Not measured | 1275 | Only Mistral Large 4 measured |
| Design Arena: game dev | Not measured | 1226 | Only Mistral Large 4 measured |
| Design Arena: data viz | Not measured | 1274 | Only Mistral Large 4 measured |
| Design Arena: 3D | Not measured | 1253 | Only Mistral Large 4 measured |
| Design Arena: all code categories | Not measured | 1259 | Only Mistral Large 4 measured |
| Design Arena: SVG | Not measured | 1233 | Only Mistral Large 4 measured |
| Design Arena: ASCII art | Not measured | 1202 | Only Mistral Large 4 measured |
| LiveBench design | Not measured | 65.6% | Only Mistral Large 4 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index013
| Benchmark | Gemma 3 12B | Mistral Large 4 | Leader |
|---|---|---|---|
| LiveBench data analysis | Not measured | 76.5% | Only Mistral Large 4 measured |
| MedScribe (Vals AI) | Not measured | 80.4% | Only Mistral Large 4 measured |
| IFBench | 36.7% | Not measured | Only Gemma 3 12B measured |
| Long context reasoning (AA-LCR) | 8.3% | 81.3% | Mistral Large 4 |
| DTBench | 48.8% | 88.2% | Mistral Large 4 |
| LMCA | 4.5% | 35.8% | Mistral Large 4 |
| Surface Evolver Bench | Not measured | 28.1% | Only Mistral Large 4 measured |
| Furniture assembly | Not measured | 20.8% | Only Mistral Large 4 measured |
| LMArena Text | 1334 | 1431 | Mistral Large 4 |
| LMArena Text: instruction following | 1299 | 1422 | Mistral Large 4 |
| LMArena Text: creative writing | 1331 | 1363 | Mistral Large 4 |
| LMArena Vision | Not measured | 1272 | Only Mistral Large 4 measured |
| LMArena Text: English | 1349 | 1433 | Mistral Large 4 |
| LMArena Text: non-English | 1318 | 1422 | Mistral Large 4 |
| LMArena Text: Chinese | Not measured | 1492 | Only Mistral Large 4 measured |
| LMArena Text: German | 1370 | Not measured | Only Gemma 3 12B measured |
| LMArena Text: Japanese | Not measured | 1415 | Only Mistral Large 4 measured |
| LMArena Text: French | Not measured | 1492 | Only Mistral Large 4 measured |
| LMArena Text: Spanish | Not measured | 1455 | Only Mistral Large 4 measured |
| LMArena Text: Russian | 1335 | 1417 | Mistral Large 4 |
| LMArena Text: multi-turn | 1334 | 1435 | Mistral Large 4 |
| LMArena Text: longer queries | 1317 | 1428 | Mistral Large 4 |
| LMArena Text: writing and language | 1307 | 1386 | Mistral Large 4 |
| LMArena Text: entertainment and media | 1297 | 1378 | Mistral Large 4 |
| LMArena Vision: English | Not measured | 1283 | Only Mistral Large 4 measured |
| LMArena Vision: reading text | Not measured | 1284 | Only Mistral Large 4 measured |
| Neither model has a result on the other 20 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemma 3 12B | Mistral Large 4 | 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 3 12B costs less
Repository review
50K input + 3K output tokens
Gemma 3 12B costs less
Agent loop
200K input + 10K output tokens
Gemma 3 12B 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 3 12B | Mistral Large 4 |
|---|---|---|
| Maker | Mistral AI | |
| Weights | Open weights | No public weights listed |
| Released | March 13, 2025 | October 6, 2026 |
| Context window | 131K | 1.05M |
| Input $/Mtok | $0.075 | $1.36 |
| Output $/Mtok | $0.1625 | $4.18 |
| Coding index | Not benchmarked | 56.4 |
| Independent measurements behind it | 0 | 5 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemma 3 12B better than Mistral Large 4 for coding?
Only Mistral Large 4 has a coding score, 56.4 on the Vibe Coding Index; Gemma 3 12B has none, so the two cannot be ranked on coding. Pick Gemma 3 12B for output tokens 25.7x cheaper ($0.1625 vs $4.18 per million).
Which is cheaper, Gemma 3 12B or Mistral Large 4?
Gemma 3 12B costs $0.075 per million input tokens and $0.1625 per million output tokens; Mistral Large 4 costs $1.36 and $4.18. Cost per solved task is known for neither of them.
Which has the bigger context window, Gemma 3 12B or Mistral Large 4?
Mistral Large 4, with 1.05M tokens against 131K.
How many benchmarks is the Gemma 3 12B vs Mistral Large 4 comparison based on?
30 benchmarks have a result for both models as of October 21, 2026. Gemma 3 12B's coding index rests on 0 independent benchmarks and Mistral Large 4's on 5. Where only one model has a result, the page says so instead of naming a winner.