Inference
EN
Google

Gemini 3.1 Pro Preview

Gemini 3.1 Pro Preview is a model from Google. 27 benchmarks count toward its index, in 7 categories.

  • Input: audio · file · image · text · video
  • Released 2026-02
RankedEnough results for an index.
IndexAll categories, out of 100.61.5±3.6Median 51.0
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.77/sMedian 67/s
First tokenSeconds to the first output token.2.9 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$2.00batch $1.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$12.00batch $6.00 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1480±3 #20Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
google/gemini-3.1-pro-preview
Max outputMaximum output tokens per request.
66K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.20
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
4 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
DatabricksCloud$2.00$12.001.05M
Google AI StudioMaker$2.00$12.001.05M
Google Vertex AIMaker$2.00$12.001.05M
OpenRouterRouter$2.00$12.001.05M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.7 counted52.5116 / 218
CodingCode writing and repair.8 counted58.877 / 209
ReasoningLogic problems and puzzles.4 counted69.821 / 227
MultimodalTasks with images and text.1 counted69.9—
KnowledgeFacts and expert knowledge.3 counted65.045 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted86.6—
MathMath problems.3 counted60.448 / 168

Results

27 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
GPQA Diamond · Vals AIKnowledge95.5% ±1.01 / 135
AIMEMath98.1% ±0.31 / 95
MMLU ProKnowledge91.0% ±0.34 / 135
ARC-AGI-1 (semi-private)Reasoning98.0%4 / 126
LiveCodeBenchCoding88.5% ±0.96 / 142
LiveBench Instruction FollowingInstructionAbove fitFar above this model's index.79.1%3 / 66
Terminal-Bench 2.0 · Vals AIAgentic67.4% ±5.04 / 66
MMMU ProMultimodal88.2% ±0.810 / 92
ARC-AGI-2 (semi-private)Reasoning77.1%21 / 131
LiveBench LanguageKnowledge85.4%12 / 66
LiveBench Data AnalysisReasoning78.5%23 / 80
SWE-benchCoding78.8% ±1.828 / 87
Terminal-Bench 2.1 · Vals AIAgentic70.8% ±1.127 / 75
LiveBench MathematicsMath91.0%28 / 66
LiveBench ReasoningReasoning84.0%42 / 80
IOICoding51.8% ±2.624 / 45
τ²-bench BankingAgentic26.0%18 / 28
Vibe Code Bench v1.1Coding32.0% ±4.380 / 123
LiveBench CodingCoding76.5%43 / 66
ARC-AGI-3 (semi-private)Reasoning0.4%18 / 26
Code MigrationCoding17.3% ±3.954 / 77
LiveBench Long AgenticAgentic23.9%37 / 51
ProofBench v1.1Math26.0% ±4.440 / 54
LiveBench ResearchAgentic65.7%40 / 52
LiveBench Coding 2.0Coding66.7%41 / 52
Agent Arena steerabilityAgentic−8.243 / 54
LiveBench Agentic CodingAgentic44.1%54 / 66
LiveBench DesignAgentic61.6%45 / 52
Agent Arena task outcomeAgentic−11.347 / 54
Terminal-Bench 4.0Agentic2.5% ±0.554 / 61
LiveBench Browser UseAgentic4.0%47 / 49
Agent Arena command recoveryAgentic−23.349 / 50
Vibe Code Bench 1-100Coding6.7% ±2.426 / 26
MedCodeHealthcare59.1% ±2.02 / 106
MedQAHealthcare96.4% ±0.22 / 95
LegalBenchLegal87.4% ±0.34 / 149
MortgageTaxFinance69.4% ±0.96 / 97
Finance Agent v1.1Finance59.7% ±2.87 / 51
ARC-AGI-2 (public eval)Reasoning88.1%18 / 122
ARC-AGI-1 (public eval)Reasoning97.2%21 / 116
CaseLaw v2Legal64.8% ±1.212 / 53
CorpFin v2Finance64.5% ±0.935 / 131
TaxEval v2Finance72.9% ±0.956 / 143
ProgramBenchCoding0.0%34 / 65
MedScribeHealthcare76.1% ±1.977 / 109
EMBFinance52.6% ±3.054 / 75
Legal Research BenchLegal20.7% ±2.867 / 90
Finance Agent (v2)Finance43.0% ±1.259 / 79
Public Benefits Bench v1Public benefits50.9% ±1.310 / 13
SAGEEducation83.2% ±2.023 / 29
Tax Agent BenchFinance9.3% ±2.168 / 82
MysteryMechanismScience20.7% ±2.725 / 30
Terminal-Bench-ScienceScience1.4% ±1.437 / 44
Public Benefits Bench v1.1Public benefits53.8% ±1.343 / 50
SRE BenchOther0.0%33 / 36
BioMysteryBenchScience60.0% ±2.928 / 30

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseVals AINo licence statedARC Prize FoundationTerms restrict reuseLiveBenchNo licence statedSierra ResearchMITLMArenaCC BY 4.0models.devMIT

History

5 releases
Index62.5 → 61.5
23 Oct 202612 Nov 2026
Output / 1M$12.00 → $12.00
23 Oct 202612 Nov 2026

Questions

How much does Gemini 3.1 Pro Preview cost?
$2.00 per 1M input tokens, $12.00 per 1M output tokens.
How good is Gemini 3.1 Pro Preview?
Its index is 61.5 out of 100. That is place 66 of 283 ranked models.
What is Gemini 3.1 Pro Preview best at?
Reasoning: 69.8 out of 100, place 21 of 227.
How fast is Gemini 3.1 Pro Preview?
77 output tokens per second, 2.9 s to the first token.
What is the context window of Gemini 3.1 Pro Preview?
1.05M tokens.
When was Gemini 3.1 Pro Preview released?
February 2026.

Badge

Gemini 3.1 Pro Preview on Inference360: place 66 of 283, index 61.5
HTML
<a href="https://inference360.ai/models/google-gemini-3-1-pro-preview"><img src="https://inference360.ai/badges/google-gemini-3-1-pro-preview.svg" alt="Gemini 3.1 Pro Preview on Inference360: place 66 of 283, index 61.5" height="20"></a>
Markdown
[![Gemini 3.1 Pro Preview on Inference360: place 66 of 283, index 61.5](https://inference360.ai/badges/google-gemini-3-1-pro-preview.svg)](https://inference360.ai/models/google-gemini-3-1-pro-preview)

Compare with

DeepSeek V4 Flash 073161.9GPT-5.3 Codex62.0Mistral Large 460.7GLM-5.260.6

Lower price

dots3-note Preview63.4 · FreeGLM-5.3-Flash63.1 · $0.078/1MLing 3.1 Flash67.2 · $0.135/1M

More from Google

All models
Gemini 3.1 Pro62.3Gemini 3.1 Pro Preview Custom ToolsUnrankedNano Banana 2 (Gemini 3.1 Flash Image Preview)UnrankedGemini 3 Pro Deep ThinkUnrankedGemini 3.1 Flash-Lite51.9