Gemini 3.1 Pro Preview

February 19, 2026Last synced Aug 1, 2026, 12:19 AM1049K context

Human Frontier

90.6

Human-calibrated frontier signal, backed by Arena-style preference evidence and separate from raw AgMoBench benchmark composite scores.

Human-preference signal trails the leading Arena frontierBenchmarks overstate practical quality

Blended Price

$4.50/M

Input Price

$2.00/M

Output Price

$12.00/M

Speed

134 tok/s

TTFT

21.37s

Benchmark Scores

Intelligence Index

Performance Radar

Gemini 3.1 Pro Preview

Field Average

External Benchmarks

AA-Omniscience Accuracy

How Gemini 3.1 Pro Preview Compares

Axes

X Axis

Y Axis

Bubble Size

Filters

Blended Price (USD)$0.00 – $37.5

AgMoBench Overall≥ 1.8

Providers

ai2ai21-labsalibabaanthropicarceeawsazurebaidu

Show quadrants

anthropic

openai

mistral

azure

cohere

meta

ai21-labs

inclusionai

google

ai2

ibm

alibaba

upstage

nvidia

aws

kimi

nous-research

thinking-machines

minimax

stepfun

tencent

deepseek

zai

multiversecomputing

xiaomi

xai

deepcogito

sarvam

kwaikat

baidu

swiss-ai-initiative

bytedance_seed

inception

reka-ai

arcee

perplexity

Bubble size = Context Window

8191.00

2000000.00

Compare with other models

GPQA Diamond

94.1 / 100

LiveCodeBench

— / 100

Terminal-Bench Hard

53.8 / 100

τ²-Bench

95.6 / 100

AA-Omniscience Hallucination Rateaa_omniscience

49.9 / 100

AA-Omniscience Indexaa_omniscience

32.9 / 100

Finance Agent v1.1anthropic_launch

59.7 / 100

HLE (with tools)anthropic_launch

51.4 / 100

ARC-AGI-1arc_agi

98.0 / 100

ARC-AGI-1 Cost per Taskarc_agi

0.5

ARC-AGI-2arc_agi

77.1 / 100

ARC-AGI-2 Cost per Taskarc_agi

1.0

ARC-AGI-1benchmark_matrix

90.0 / 100

ARC-AGI-2benchmark_matrix

77.1 / 100

AA Intelligence Index (Matrix)benchmark_matrix

57.0

AIME 2024benchmark_matrix

98.0

AIME 2025 (Matrix)benchmark_matrix

100.0

AIME 2026benchmark_matrix

97.0

BrowseCompbenchmark_matrix

85.9

Chatbot Arena ELO (Matrix)benchmark_matrix

1490.0

Codeforces Ratingbenchmark_matrix

2700.0

CritPtbenchmark_matrix

17.7

GPQA Diamond (Matrix)benchmark_matrix

94.3

HLE (Matrix)benchmark_matrix

44.4

HumanEvalbenchmark_matrix

95.0

IFEvalbenchmark_matrix

89.2

LiveCodeBench (Matrix)benchmark_matrix

82.0

MATH-500 (Matrix)benchmark_matrix

98.5

MathArena Apex 2025benchmark_matrix

33.5

MathVisionbenchmark_matrix

69.8

MMLUbenchmark_matrix

92.6

MMLU-Pro (Matrix)benchmark_matrix

89.5

MMMU-Probenchmark_matrix

80.5

MRCR v2benchmark_matrix

84.9

OSWorld (Unversioned Matrix Import)benchmark_matrix

72.0 / 100

SciCode (Matrix)benchmark_matrix

58.9

SimpleQAbenchmark_matrix

72.1

SWE-bench Probenchmark_matrix

54.2

Terminal-Bench 2.0benchmark_matrix

68.5

browsecompbenchmark_matrix

85.9

FrontierMathbenchmark_matrix

40.0 / 100

MMMUbenchmark_matrix

87.5 / 100

SimpleBenchbenchmark_matrix

79.6 / 100

simpleqabenchmark_matrix

72.1

SWE-bench Verifiedbenchmark_matrix

80.6 / 100

tau-bench Retailbenchmark_matrix

90.5 / 100

AA-Omniscience Hallucination RatePredicted

93.9 / 100

Aider PolyglotPredicted

85.6 / 100

AIMEPredicted

0.3 / 30

AIME 2025Predicted

0.9 / 30

AlpacaEval 2.0 LCPredicted

14.3 / 100

AlpacaEval 2.0 RawPredicted

9.5 / 100

BFCL (Berkeley Function Calling)Predicted

56.8

BigCodeBench CompletePredicted

37.4 / 100

BigCodeBench InstructPredicted

29.2 / 100

AA Long Context Reasoning (Matrix)Predicted

76.8

Arena-Hard AutoPredicted

74.5

BRUMO 2025Predicted

99.8

CMIMC 2025Predicted

38.4

HMMT Feb 2025Predicted

97.9

HMMT Nov 2025Predicted

93.3

IFBench (Matrix)Predicted

72.8

IMO 2025Predicted

46.9

SMT 2025Predicted

95.5

Tau-Bench Telecom (Matrix)Predicted

99.0

Terminal-Bench 1.0Predicted

55.9

USAMO 2025Predicted

69.2

Video-MMUPredicted

94.0

Aider PolyglotPredicted

0.3

BIG-Bench HardPredicted

3.0

CAD-EvalPredicted

2.3

CyBenchPredicted

1.1

DeepResearchBenchPredicted

0.5

FictionLiveBenchPredicted

0.7

GeoBenchPredicted

0.0

GSM8K (Epoch)Predicted

1.8

HellaSwagPredicted

0.0

Lech Mazur WritingPredicted

8.8

METR Time HorizonsPredicted

57.3

The Agent CompanyPredicted

56.4

TriviaQAPredicted

27.7

WinoGrandePredicted

1.6

GAIA Level 1Predicted

84.3

GAIA Level 2Predicted

69.0

GAIA Level 3Predicted

89.4

GAIAPredicted

71.0 / 100

LiveCodeBenchPredicted

0.9 / 100

LongBench v2 EasyPredicted

56.7

LongBench v2 HardPredicted

28.7

LongBench v2Predicted

25.3 / 100

MATH-500Predicted

0.9 / 100

MathVistaPredicted

57.7 / 100

MLE-bench (Original)Predicted

67.1 / 100

MMLU ProPredicted

0.9 / 100

MMTU Table UnderstandingPredicted

73.9 / 100

MT-BenchPredicted

7.8 / 10

NoLiMa (NIAH)Predicted

93.2 / 100

OCRBench v2Predicted

86.0 / 100

Open LLM AveragePredicted

18.4 / 100

Open LLM: BBHPredicted

49.3 / 100

Open LLM: GPQAPredicted

27.4 / 100

Open LLM: IFEvalPredicted

24.7 / 100

Open LLM: MATH Level 5Predicted

53.9 / 100

Open LLM: MMLU-PROPredicted

33.3 / 100

Open LLM: MUSRPredicted

39.8 / 100

RE-BenchPredicted

100.0

SWE-bench LitePredicted

65.7 / 100

WebArenaPredicted

4.3 / 100

WildBenchPredicted

34.6

browsecompbrowsecomp

85.9

BullshitBenchbullshitbench

37.0 / 100

Arena ELO: Codingchatbot_arena

1447.0

Arena ELO: Creative Writingchatbot_arena

1490.0

arena_elo_documentchatbot_arena

1445.0

Chatbot Arena ELOchatbot_arena

1486.0

arena_elo_visionchatbot_arena

1278.0

Ale Benchepoch_ai

465.1

Algotuneepoch_ai

2.0

Apex Agentsepoch_ai

3.3

Arc Agi 2epoch_ai

0.8

Blueprint Bench 2epoch_ai

0.3

Chess Puzzlesepoch_ai

0.6

Cl Benchepoch_ai

0.2

Cl Bench Lifeepoch_ai

0.2

DeepSWE (Epoch Import, Version Unverified)epoch_ai

81.4 / 100

Enigma Evalepoch_ai

0.2

Epoch Capabilities Indexepoch_ai

154.9

Exploitbenchepoch_ai

3.2

Forecastbenchepoch_ai

1.0

Frontiercodeepoch_ai

0.0

Frontiersweepoch_ai

11.0

METR Time Horizonsepoch_ai

89.8

Mystery Game Puzzlesepoch_ai

0.3

OTIS Mock AIME 2024–2025epoch_ai

1.0

Posttrainbenchepoch_ai

0.2

Proofbenchepoch_ai

0.3

SimpleQA Verified (Epoch)epoch_ai

0.8

Vending Bench 2epoch_ai

911.2

WebDev Arenaepoch_ai

1446.5

EQ-Bench 3eqbench

1302.0

FrontierMathfrontiermath

36.9 / 100

IDP Coreidp-leaderboard

89.6 / 100

IDP OlmOCRidp-leaderboard

69.8 / 100

IDP OmniDocidp-leaderboard

85.3 / 100

IDP Overallidp-leaderboard

81.6 / 100

LegalBenchlegalbench

87.4 / 100

LiveBench Codinglivebench

76.5 / 100

LiveBench Data Analysislivebench

78.5 / 100

LiveBench Instruction Followinglivebench

79.1 / 100

LiveBench Languagelivebench

85.4 / 100

LiveBench Mathlivebench

91.0 / 100

LiveBench Overalllivebench

79.9 / 100

LiveBench Reasoninglivebench

84.0 / 100

AIME 2026matharena

98.3

MedQA (USMLE)medqa

96.4

OpenRouter Token Volumeopenrouter_rankings

369000000000.0

OpenRouter Weekly Trendopenrouter_rankings

23.0

MultiChallengeseal

71.4

SWE-Bench Pro (Private)seal

32.2

SWE-Bench Pro (Public)seal

46.1

SimpleBenchsimplebench

79.6 / 100

SWE-bench Verifiedswebench

80.6 / 100

Vectara Factual Consistencyvectara_hallucination

89.6 / 100

Vectara Hallucination Ratevectara_hallucination

10.4 / 100

WeirdMLweirdml

72.1 / 100

ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 3.1 Pro (Preview) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-3-1-pro-preview

View run source

Source metadata

{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-3-1-pro-preview",
  "modelNotes": null,
  "modelsEtag": "W/\"2fc19b6caba8b52765ca17eab04fc437\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9485c7ef206d9b12cdfdec6a631df56b\"",
  "upstreamModelId": "google-gemini-3-1-pro-preview",
  "environmentCount": 25,
  "modelReleaseDate": "2026-03-05T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 2213.53,
  "providerModelIdPublished": false
}

Evidence observed 2026-07-31

0.4 / 100

ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 3.1 Pro (Preview) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-3-1-pro-preview

View run source

Source metadata

{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-3-1-pro-preview",
  "modelNotes": null,
  "modelsEtag": "W/\"2fc19b6caba8b52765ca17eab04fc437\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9485c7ef206d9b12cdfdec6a631df56b\"",
  "upstreamModelId": "google-gemini-3-1-pro-preview",
  "environmentCount": 25,
  "modelReleaseDate": "2026-03-05T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 2213.53,
  "providerModelIdPublished": false
}

Evidence observed 2026-07-31

0.4 / 100

ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 3.1 Pro (Preview) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-3-1-pro-preview

View run source

Source metadata

{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-3-1-pro-preview",
  "modelNotes": null,
  "modelsEtag": "W/\"cfc9b87c645d3ef2103e359fd0f9f521\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"0d65f6e0efe0d5cae15e8e2fe5a0994b\"",
  "upstreamModelId": "google-gemini-3-1-pro-preview",
  "environmentCount": 25,
  "modelReleaseDate": "2026-03-05T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 2213.53,
  "providerModelIdPublished": false
}

Evidence observed 2026-07-30

0.4 / 100

ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 3.1 Pro (Preview) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-3-1-pro-preview

View run source

Source metadata

{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-3-1-pro-preview",
  "modelNotes": null,
  "modelsEtag": "W/\"bbdf5725ef7a703a7f0970e913262b6f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6529094085f9ca7ce5f3c2e1957d3b9c\"",
  "upstreamModelId": "google-gemini-3-1-pro-preview",
  "environmentCount": 25,
  "modelReleaseDate": "2026-03-05T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 2213.53,
  "providerModelIdPublished": false
}

Evidence observed 2026-07-24

0.4 / 100

Senior SWE-Bench v2026.06senior_swe_bench

v2026.06 · official-100-task · Gemini 3.1 Pro (high) · harness mini-swe-agent · effort high · 1 attempt · pass@1 · harbor:snorkel-ai/senior-swe-bench-v2026.06 · provider google/gemini-3.1-pro-preview · record gemini-3-1-pro-mini-swe · run e7a084b8bf4fbf7e4723f412d7fc0b621bc5db20fd0e61b5014f602fa22d436e

View run source

Source metadata

{
  "taskCount": 100,
  "agentGroup": "mini-swe-agent:gemini-3.1-pro-preview",
  "basicPassAt1": 9.473684210526317,
  "basicPassAt3": 24.210526315789473,
  "datasetPackage": "snorkel-ai/senior-swe-bench-v2026.06",
  "publicTaskCount": 50,
  "tastefulPassAt1": 2.1052631578947367,
  "tastefulPassAt3": 8.421052631578947,
  "agentDescription": "Google Gemini 3.1 Pro via mini-swe-agent harness (Portkey→Google)",
  "privateTaskCount": 50,
  "upstreamProvider": "google",
  "upstreamRunCount": 50,
  "averageAgentSteps": 107.73684210526316,
  "averageOutputTokens": 17277.894736842107,
  "averageTotalCostUsd": 1.3306002509473684,
  "averageOutputCostUsd": 0.2073347368421053,
  "basicAllThreePassRate": 5.263157894736842,
  "tastefulAllThreePassRate": 0
}

Evidence observed 2026-07-29

2.1 / 100

Senior SWE-Bench v2026.06senior_swe_bench

v2026.06 · official-100-task · Gemini 3.1 Pro (high) · harness mini-swe-agent · effort high · 1 attempt · pass@1 · harbor:snorkel-ai/senior-swe-bench-v2026.06 · provider google/gemini-3.1-pro-preview · record gemini-3-1-pro-mini-swe · run 3018be1dee8e4b352d3aa9260d805ee298040ab383dec290a16294f9afd8472d

View run source

Source metadata

{
  "taskCount": 100,
  "agentGroup": "mini-swe-agent:gemini-3.1-pro-preview",
  "basicPassAt1": 26.136363636363637,
  "datasetPackage": "snorkel-ai/senior-swe-bench-v2026.06",
  "publicTaskCount": 50,
  "tastefulPassAt1": 6.8181818181818175,
  "agentDescription": "Google Gemini 3.1 Pro via mini-swe-agent harness (Portkey→Google)",
  "privateTaskCount": 50,
  "upstreamProvider": "google",
  "upstreamRunCount": 50,
  "averageAgentSteps": 89.52272727272727,
  "averageOutputTokens": 19415.81818181818,
  "averageTotalCostUsd": 0.8407021872727273,
  "averageOutputCostUsd": 0.23298981818181816
}

Evidence observed 2026-06-30

6.8 / 100

SWE-rebenchswe_rebench

vcontinuous-leaderboard · tasks-created:2026-02-01:2026-05-15 · Gemini 3.1 Pro Preview · harness swe-rebench-react@tools · 5 attempts · pass@1 · swe-rebench:standardized:128k-or-model-limit · provider Gemini 3.1 Pro Preview__tools · record Gemini 3.1 Pro Preview__tools:1769904000000:1778803200000

View run source

Source metadata

{
  "passAt5": 69.46107784431138,
  "developer": "google",
  "instanceType": "model",
  "contextPolicy": "128k unless the model supports less",
  "attemptsPerTask": 5,
  "totalTokenUsage": 1502695.5944015582,
  "modelReleaseDate": "2026-02-19",
  "scoreStandardError": 0.9239067816830353,
  "cachedTokenPercentage": 80.60600910081673,
  "modelReleaseTimestamp": 1771459200000,
  "standardContextTokens": 128000,
  "averageInstanceCostUsd": 0.7208578944116585,
  "potentialContamination": true,
  "aggregateTaskRangeTimestamp": {
    "to": 1778803200000,
    "from": 1769904000000
  },
  "publishedTaskRangeTimestamp": {
    "to": 1778803200000,
    "from": 1769904000000
  }
}

Evidence observed 2026-05-15

54.9 / 100

Terminal-Bench 2.1terminal_bench

v2.1 · official-leaderboard · Gemini 3.1 Pro (high) · harness terminus-2@2.0.0 · effort high · 5 attempts · harbor:terminal-bench/terminal-bench-2-1 · provider gemini/gemini-3.1-pro-preview · record leaderboard/submissions/2026-05-05-gemini-gemini-3-1-pro-preview-high-terminus-2.json · run 42cd19c9-42ad-5d79-b033-adf4f879423d

View run source

Source metadata

{
  "passAt2": 0.7775,
  "passAt3": 0.8225,
  "passAt4": 0.8472,
  "passAt5": 0.8652,
  "taskCount": 89,
  "actualTokens": {
    "output": 12935831,
    "cachedInput": 96199940,
    "uncachedInput": 27525560
  },
  "datasetPackage": "terminal-bench/terminal-bench-2-1",
  "leaderboardUrl": "https://www.tbench.ai/leaderboard/terminal-bench/2.1",
  "pullRequestUrl": "https://github.com/harbor-framework/terminal-bench-2-1/pull/69",
  "agentDisplayUrl": "https://www.tbench.ai/news/terminus",
  "modelDisplayUrl": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
  "actualTrialCount": 445,
  "agentOrganization": "Terminal-Bench",
  "modelOrganization": "Google",
  "actualTotalCostUsd": 229.99,
  "rewardHacksPercent": 0.45,
  "accuracyStandardError": 1.65,
  "submissionDownloadUrl": "https://raw.githubusercontent.com/harbor-framework/terminal-bench-2-1/main/leaderboard/submissions/2026-05-05-gemini-gemini-3-1-pro-preview-high-terminus-2.json",
  "disqualifiedTrialCount": 2,
  "expectedAttemptsPerTask": 5,
  "sourceFilterReasoningEffort": "high",
  "actualAverageTrialDurationSeconds": 672.8
}

Evidence observed 2026-05-05

65.6 / 100

Terminal-Bench 2.1terminal_bench

v2.1 · official-leaderboard · Gemini 3.1 Pro (high) · harness gemini-cli@0.40.0 · effort high · 5 attempts · harbor:terminal-bench/terminal-bench-2-1 · provider gemini/gemini-3.1-pro-preview · record leaderboard/submissions/2026-05-05-gemini-gemini-3-1-pro-preview-high-gemini-cli.json · run 42cd19c9-42ad-5d79-b033-adf4f879423d

View run source

Source metadata

{
  "passAt2": 0.782,
  "passAt3": 0.8315,
  "passAt4": 0.8584,
  "passAt5": 0.8764,
  "taskCount": 89,
  "actualTokens": {
    "output": 6123596,
    "cachedInput": 366583353,
    "uncachedInput": 44843158
  },
  "datasetPackage": "terminal-bench/terminal-bench-2-1",
  "leaderboardUrl": "https://www.tbench.ai/leaderboard/terminal-bench/2.1",
  "pullRequestUrl": "https://github.com/harbor-framework/terminal-bench-2-1/pull/68",
  "agentDisplayUrl": "https://github.com/google-gemini/gemini-cli",
  "modelDisplayUrl": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
  "actualTrialCount": 445,
  "agentOrganization": "Google",
  "modelOrganization": "Google",
  "actualTotalCostUsd": 236.49,
  "rewardHacksPercent": 0.22,
  "accuracyStandardError": 1.67,
  "submissionDownloadUrl": "https://raw.githubusercontent.com/harbor-framework/terminal-bench-2-1/main/leaderboard/submissions/2026-05-05-gemini-gemini-3-1-pro-preview-high-gemini-cli.json",
  "disqualifiedTrialCount": 1,
  "expectedAttemptsPerTask": 5,
  "sourceFilterReasoningEffort": "high",
  "actualAverageTrialDurationSeconds": 652.9
}

Evidence observed 2026-05-05

65.8 / 100