Gemini 3.1 Pro Preview vs Gemini 4 Carbon
Gemini 4 Carbon leads Gemini 3.1 Pro Preview on coding, 86.6 to 32.3 on the Vibe Coding Index, and Gemini 4 Carbon costs 1.2x less per output token ($10.00 vs $12.00 per million). Updated November 14, 2026; 56 benchmarks have a result for both.
vs boards won of 56 shared
Gemini 3.1 Pro Preview or Gemini 4 Carbon: which to pick
Pick Gemini 4 Carbon for coding: it scores 86.6 on the Vibe Coding Index against 32.3 for Gemini 3.1 Pro Preview.
Pick Gemini 3.1 Pro Preview for work measured by Long context reasoning (AA-LCR), where it leads.
Model AGemini 3.1 Pro PreviewGoogle32.3/1004 independent benchmarks$0.99 per solved taskModel BGemini 4 CarbonGoogle86.6/1004 independent benchmarksNo cost per solved task
- Shared results
- 56
- Gemini 3.1 Pro Preview only
- 87
- Gemini 4 Carbon only
- 17
- 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.
Coding
Vibe Coding Index, six public coding benchmarks
Gemini 4 Carbon
Gemini 4 Carbon scores 86.6 against 32.3.
4 benchmarks behind each score
Intelligence
AA Intelligence Index
No call
Only Gemini 4 Carbon has a AA Intelligence Index result, 60.0.
Directional only
Agentic work
AA Agentic Index
No call
Only Gemini 3.1 Pro Preview has a AA Agentic Index result, 8.2.
Directional only
Reasoning
GPQA Diamond
No call
Only Gemini 3.1 Pro Preview has a GPQA Diamond result, 95.5%.
Directional only
Math
AIME
No call
Only Gemini 3.1 Pro Preview has a AIME result, 98.1%.
Directional only
Knowledge
MMLU Pro
No call
Only Gemini 3.1 Pro Preview has a MMLU Pro result, 91.0%.
Directional only
Web design
Design Arena: websites
Gemini 4 Carbon
Gemini 4 Carbon leads, 1319 against 1226.
Same board, both measured
General
Vals Index
No call
Only Gemini 4 Carbon has a Vals Index result, 70.5%.
Directional only
Long documents
Listed context window
No call
Both take 1.05M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Gemini 4 Carbon
Gemini 3.1 Pro Preview $0.0080, Gemini 4 Carbon $0.0070 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Gemini 3.1 Pro Preview has the number, $0.99.
Directional only
At a glance
- Coding index
- Gemini 3.1 Pro Preview 32.3 from 4 benchmarks, Gemini 4 Carbon 86.6 from 4 benchmarks.
- Widest category
- Agentic work, 46.8 points apart on average over 6 shared benchmarks.
- Widest single gap
- Vending-Bench 2: Gemini 3.1 Pro Preview $911 against Gemini 4 Carbon $16,484, 94.3 points apart on a 0 to 100 scale.
- Output price
- Gemini 3.1 Pro Preview $12.00 against Gemini 4 Carbon $10.00 per million output tokens, 1.2x apart.
- Context window
- Both take 1.05M 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 | Gemini 3.1 Pro Preview | Bars | Gemini 4 Carbon | Ties |
|---|---|---|---|---|
| Coding | 0 | 9 | ||
| Agentic work | 0 | 6 | ||
| Reasoning | 0 | 11 | ||
| Math | 0 | 5 | ||
| Knowledge | 0 | 6 | ||
| Web design | 0 | 5 | ||
| General | 1 | 13 | ||
| All shared boards | 1 | 55 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index09
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 91.0% | Only Gemini 4 Carbon measured |
| LMArena WebDev | 1446 | 1768 | Gemini 4 Carbon |
| AA Coding Index | 68.8 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LiveCodeBench (Vals AI) | 88.5% | Not measured | Only Gemini 3.1 Pro Preview measured |
| SWE-bench Verified (Vals AI) | 78.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Terminal-Bench 4 (Vals AI) | 2.5% | 63.1% | Gemini 4 Carbon |
| Vibe Code Bench (Vals AI) | 32.0% | 89.1% | Gemini 4 Carbon |
| Code migration (Vals AI) | 17.3% | Not measured | Only Gemini 3.1 Pro Preview measured |
| ProgramBench (Vals AI) | 0.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| IOI (Vals AI) | 51.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Terminal-Bench 2.1 (Vals AI) | 70.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Vibe Code Bench 1-100 (Vals AI) | 6.7% | Not measured | Only Gemini 3.1 Pro Preview measured |
| SciCode | 58.7% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Terminal-Bench Hard (Artificial Analysis run) | 53.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Terminal-Bench 4.0 (Artificial Analysis run) | 4.0% | 65.2% | Gemini 4 Carbon |
| FrontierCode 2.0 | Not measured | 40.6% | Only Gemini 4 Carbon measured |
| Terminal-Bench | 80.2% | Not measured | Only Gemini 3.1 Pro Preview measured |
| FrontierSWE | Not measured | 61.2% | Only Gemini 4 Carbon measured |
| GSO-Bench | 22.6% | Not measured | Only Gemini 3.1 Pro Preview measured |
| WeirdML | 72.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| MirrorCode | 8.9% | Not measured | Only Gemini 3.1 Pro Preview measured |
| ALE-Bench | 1160.6 | 2374.7 | Gemini 4 Carbon |
| LMArena Text: coding | 1483 | 1564 | Gemini 4 Carbon |
| LMArena Text: software and IT | 1486 | 1552 | Gemini 4 Carbon |
| LMArena WebDev: React | 1437 | 1792 | Gemini 4 Carbon |
| LMArena WebDev: HTML | 1489 | 1770 | Gemini 4 Carbon |
| LMArena WebDev: from an image | 1480 | Not measured | Only Gemini 3.1 Pro Preview measured |
| SciCode | 14.3% | Not measured | Only Gemini 3.1 Pro Preview measured |
| GameBench 2 | 65.0 | Not measured | Only Gemini 3.1 Pro Preview measured |
| CursorBench 4.0 | Not measured | 60.4% | Only Gemini 4 Carbon measured |
| Neither model has a result on the other 22 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 60.0 | Only Gemini 4 Carbon measured |
| Epoch Capabilities Index | Not measured | 169.8 | Only Gemini 4 Carbon measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| AA Agentic Index | 8.2 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: agents building web apps | 1102 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: agents building full-stack apps | 1042 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: agents building mobile apps | 1086 | 1284 | Gemini 4 Carbon |
| Design Arena: agents building Android apps | 1035 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: agents building Godot games | 1143 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: agents building games | 1057 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: HTML slides | 1048 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: PowerPoint slides | 1035 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: spreadsheets | 944 | 1224 | Gemini 4 Carbon |
| LiveBench research | Not measured | 92.5% | Only Gemini 4 Carbon measured |
| LiveBench browser | Not measured | 80.0% | Only Gemini 4 Carbon measured |
| LiveBench long agentic | Not measured | 72.7% | Only Gemini 4 Carbon measured |
| SREBench (Vals AI) | 0.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Finance Agent (Vals AI) | 43.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Tax Agent Bench (Vals AI) | 9.3% | 51.8% | Gemini 4 Carbon |
| Legal Research Bench (Vals AI) | 20.7% | 55.8% | Gemini 4 Carbon |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 17.5% | Only Gemini 4 Carbon measured |
| Excel Modeling Benchmark (Vals AI) | 52.6% | Not measured | Only Gemini 3.1 Pro Preview measured |
| BioMysteryBench (Vals AI) | 60.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Terminal-Bench-Science (Vals AI) | 1.4% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Vals Science Index | 27.4% | Not measured | Only Gemini 3.1 Pro Preview measured |
| τ²-Bench Telecom | 95.6% | Not measured | Only Gemini 3.1 Pro Preview measured |
| APEX-Agents | 35.1% | 84.2% | Gemini 4 Carbon |
| METR time horizons | 77.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| BALROG | 57.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| DeepResearch Bench | 47.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| PostTrainBench | 18.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Vending-Bench 2 | $911 | $16,484 | Gemini 4 Carbon |
| MCP Atlas | 78.2% | Not measured | Only Gemini 3.1 Pro Preview measured |
| τ³-Banking | 26.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Neither model has a result on the other 13 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index011
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| GPQA Diamond | 95.5% | Not measured | Only Gemini 3.1 Pro Preview measured |
| LiveBench reasoning | Not measured | 90.6% | Only Gemini 4 Carbon measured |
| MysteryMechanism (Vals AI) | 20.7% | Not measured | Only Gemini 3.1 Pro Preview measured |
| MMMU Pro (Vals AI) | 88.2% | Not measured | Only Gemini 3.1 Pro Preview measured |
| GPQA Diamond (Artificial Analysis run) | 94.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Humanity's Last Exam (Artificial Analysis run) | 47.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| HLE-Diamond (Artificial Analysis run) | 31.4% | 69.1% | Gemini 4 Carbon |
| CritPt v2, hard subset (Artificial Analysis run) | Not measured | 69.0% | Only Gemini 4 Carbon measured |
| GPQA Diamond (Epoch AI run) | 94.4% | 96.5% | Gemini 4 Carbon |
| ARC-AGI | 98.0% | 98.5% | Gemini 4 Carbon |
| ARC-AGI-2 | 77.1% | 94.2% | Gemini 4 Carbon |
| SimpleBench | 79.6% | 92.6% | Gemini 4 Carbon |
| Chess puzzles | 51.0% | 78.0% | Gemini 4 Carbon |
| Mystery game puzzles | 34.0% | 73.0% | Gemini 4 Carbon |
| EBR-bench | 14.3% | 87.6% | Gemini 4 Carbon |
| LMArena Text: hard prompts | 1485 | 1551 | Gemini 4 Carbon |
| LMArena Text: expert | 1483 | 1582 | Gemini 4 Carbon |
| LMArena Text: hard prompts in English | 1480 | 1545 | Gemini 4 Carbon |
| LMArena Vision: diagrams | 1311 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: homework | 1317 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Neither model has a result on the other 6 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| AIME | 98.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| OTIS Mock AIME | 95.6% | 100.0% | Gemini 4 Carbon |
| FrontierMath, tiers 1 to 3 | 59.6% | 90.5% | Gemini 4 Carbon |
| FrontierMath, tier 4 | 26.8% | 95.1% | Gemini 4 Carbon |
| ProofBench | 26.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: math | 1485 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: mathematical work | 1476 | 1538 | Gemini 4 Carbon |
| MathArena | 28.1% | 85.4% | Gemini 4 Carbon |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| MMLU Pro | 91.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| LegalBench (Vals AI) | 87.4% | Not measured | Only Gemini 3.1 Pro Preview measured |
| CaseLaw (Vals AI) | 64.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| CorpFin (Vals AI) | 64.5% | Not measured | Only Gemini 3.1 Pro Preview measured |
| TaxEval (Vals AI) | 72.9% | Not measured | Only Gemini 3.1 Pro Preview measured |
| MortgageTax (Vals AI) | 69.4% | Not measured | Only Gemini 3.1 Pro Preview measured |
| MedQA (Vals AI) | 96.4% | Not measured | Only Gemini 3.1 Pro Preview measured |
| MedCode (Vals AI) | 59.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| AA-Omniscience | Not measured | 53.5 | Only Gemini 4 Carbon measured |
| SimpleQA Verified | 73.5% | 77.7% | Gemini 4 Carbon |
| GeoBench | 84.0% | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: science | 1496 | 1535 | Gemini 4 Carbon |
| LMArena Text: medicine and healthcare | 1486 | 1525 | Gemini 4 Carbon |
| LMArena Text: legal and government | 1486 | 1542 | Gemini 4 Carbon |
| LMArena Text: business and finance | 1460 | 1520 | Gemini 4 Carbon |
| FACTS Benchmark Suite | 69.1% | 78.4% | Gemini 4 Carbon |
| PRBench Legal | Not measured | 53.3% | Only Gemini 4 Carbon measured |
| Neither model has a result on the other 4 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| Design Arena: websites | 1226 | 1319 | Gemini 4 Carbon |
| Design Arena: UI components | 1215 | 1357 | Gemini 4 Carbon |
| Design Arena: game dev | 1191 | 1326 | Gemini 4 Carbon |
| Design Arena: data viz | 1206 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: 3D | 1214 | 1434 | Gemini 4 Carbon |
| Design Arena: all code categories | 1220 | 1348 | Gemini 4 Carbon |
| Design Arena: SVG | 1275 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Design Arena: ASCII art | 1265 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LiveBench design | Not measured | 80.2% | Only Gemini 4 Carbon measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index113
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | Leader |
|---|---|---|---|
| Vals Index | Not measured | 70.5% | Only Gemini 4 Carbon measured |
| LiveBench data analysis | Not measured | 86.5% | Only Gemini 4 Carbon measured |
| Public Benefits Bench (Vals AI) | 53.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| SAGE (Vals AI) | 83.2% | Not measured | Only Gemini 3.1 Pro Preview measured |
| MedScribe (Vals AI) | 76.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Vals Multimodal Index | 56.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| IFBench | 77.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Long context reasoning (AA-LCR) | 82.0% | 79.3% | Gemini 3.1 Pro Preview |
| GDPval-AA | 799 | 1733 | Gemini 4 Carbon |
| DTBench | 97.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMCA | 53.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| CL-bench | 20.8% | Not measured | Only Gemini 3.1 Pro Preview measured |
| Furniture assembly | 26.7% | 77.5% | Gemini 4 Carbon |
| ExploitBench | 26.1% | Not measured | Only Gemini 3.1 Pro Preview measured |
| CL-bench Life | 16.9% | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text | 1480 | 1538 | Gemini 4 Carbon |
| LMArena Text: instruction following | 1465 | 1539 | Gemini 4 Carbon |
| LMArena Text: creative writing | 1483 | 1533 | Gemini 4 Carbon |
| LMArena Vision | 1296 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: English | 1478 | 1530 | Gemini 4 Carbon |
| LMArena Text: non-English | 1476 | 1534 | Gemini 4 Carbon |
| LMArena Text: Chinese | 1528 | 1613 | Gemini 4 Carbon |
| LMArena Text: German | 1491 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: Japanese | 1492 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: Korean | 1456 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: French | 1487 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: Spanish | 1479 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: Russian | 1497 | 1560 | Gemini 4 Carbon |
| LMArena Text: Polish | 1498 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Text: multi-turn | 1486 | 1569 | Gemini 4 Carbon |
| LMArena Text: longer queries | 1482 | 1549 | Gemini 4 Carbon |
| LMArena Text: writing and language | 1481 | 1542 | Gemini 4 Carbon |
| LMArena Text: entertainment and media | 1465 | 1527 | Gemini 4 Carbon |
| LMArena Vision: English | 1287 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: Chinese | 1362 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: reading text | 1304 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: humor | 1323 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: captioning | 1343 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: recognizing things | 1329 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Vision: creative writing | 1324 | Not measured | Only Gemini 3.1 Pro Preview measured |
| LMArena Document | 1443 | Not measured | Only Gemini 3.1 Pro Preview measured |
| MultiChallenge | 71.4% | Not measured | Only Gemini 3.1 Pro Preview measured |
| EQ-Bench 4 | 1132 | Not measured | Only Gemini 3.1 Pro Preview measured |
| Neither model has a result on the other 6 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Gemini 3.1 Pro Preview | Gemini 4 Carbon | 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
Gemini 3.1 Pro Preview$0.0080Fits in one request
Gemini 4 Carbon$0.0070Fits in one request
Gemini 4 Carbon costs less
Repository review
50K input + 3K output tokens
Gemini 3.1 Pro Preview$0.136Fits in one request
Gemini 4 Carbon$0.130Fits in one request
Gemini 4 Carbon costs less
Agent loop
200K input + 10K output tokens
Gemini 3.1 Pro Preview$0.520Fits in one request
Gemini 4 Carbon$0.500Fits in one request
Gemini 4 Carbon 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 Gemini 3.1 Pro Preview | Google Gemini 4 Carbon |
|---|---|---|
| Maker | ||
| Weights | No public weights listed | No public weights listed |
| Released | February 19, 2026 | November 10, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $2.00 | $2.00 |
| Output $/Mtok | $12.00 | $10.00 |
| Coding index | 32.3 | 86.6 |
| Independent measurements behind it | 4 | 4 |
| Cost per solved task | $0.99 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Gemini 3.1 Pro Preview better than Gemini 4 Carbon for coding?
Pick Gemini 4 Carbon for coding: it scores 86.6 on the Vibe Coding Index against 32.3 for Gemini 3.1 Pro Preview. Pick Gemini 3.1 Pro Preview for work measured by Long context reasoning (AA-LCR), where it leads.
Which is cheaper, Gemini 3.1 Pro Preview or Gemini 4 Carbon?
Gemini 3.1 Pro Preview costs $2.00 per million input tokens and $12.00 per million output tokens; Gemini 4 Carbon costs $2.00 and $10.00. Cost per solved task is known for Gemini 3.1 Pro Preview only.
Which has the bigger context window, Gemini 3.1 Pro Preview or Gemini 4 Carbon?
Neither: both take 1.05M tokens.
How many benchmarks is the Gemini 3.1 Pro Preview vs Gemini 4 Carbon comparison based on?
56 benchmarks have a result for both models as of November 14, 2026. Gemini 3.1 Pro Preview's coding index rests on 4 independent benchmarks and Gemini 4 Carbon's on 4. Where only one model has a result, the page says so instead of naming a winner.