Inference
EN
Google

Gemini 3.7 Flash

Gemini 3.7 Flash is a reasoning model from Google. 53 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Reasoning
  • Input: text · image · video · file · audio
  • Released 2026-08
RankedEnough results for an index.
IndexAll categories, out of 100.68.5±2.2Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.85/sMedian 67/s
First tokenSeconds to the first output token.2.4 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.75batch $0.375Median $0.60
Output / 1MUS dollars per 1M output tokens.$3.75batch $1.88 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1489±5 #15Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
google/gemini-3.7-flash
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.
$0.075
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
3 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Google AI StudioMaker$0.75$3.751.05M
Google Vertex AIMaker$0.75$3.751.05M
OpenRouterRouter$0.75$3.751.05M

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.16 counted65.145 / 218
CodingCode writing and repair.12 counted66.930 / 209
ReasoningLogic problems and puzzles.8 counted68.724 / 227
MultimodalTasks with images and text.4 counted67.715 / 84
KnowledgeFacts and expert knowledge.6 counted71.215 / 268
MultilingualTasks in many languages.1 counted72.3—
InstructionTasks with strict prompt rules.1 counted88.3—
MathMath problems.5 counted66.826 / 168

Results

53 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
EuroEval PolishMultilingual67.9% ±0.41 / 209
EuroEval FrenchMultilingual77.1% ±0.31 / 178
EuroEval GermanMultilingual67.3% ±0.52 / 191
EuroEval SpanishMultilingual68.3% ±0.42 / 176
EuroEval SwedishMultilingual76.6% ±0.53 / 195
EuroEval PortugueseMultilingual74.4% ±0.53 / 182
EuroEval DutchMultilingual73.1% ±0.43 / 175
EuroEval ItalianMultilingual76.9% ±0.33 / 174
Artificial Analysis GPQA DiamondKnowledge94.5%4 / 199
LiveBench Instruction FollowingInstructionAbove fitFar above this model's index.79.9%2 / 66
GPQA diamond · Epoch AIKnowledge94.8% ±1.37 / 217
LiveCodeBenchCoding88.7% ±0.95 / 142
GPQA Diamond · Vals AIKnowledge93.9% ±1.55 / 135
Artificial Analysis AnalystAgentAgentic60.0%1 / 26
Artificial Analysis MMMU-ProMultimodal85.5%5 / 113
MMLU ProKnowledge90.1% ±0.36 / 135
MMMU ProMultimodal89.0% ±0.85 / 92
Chess PuzzlesReasoning66.0% ±4.86 / 90
Artificial Analysis Omniscience AccuracyKnowledge55.3%12 / 164
Artificial Analysis Humanity's Last ExamKnowledge47.9%18 / 222
SkillsBenchCoding65.9% ±4.63 / 35
ARC-AGI-2 (semi-private)Reasoning84.6%14 / 131
SimpleQA VerifiedKnowledge69.2% ±1.512 / 91
ARC-AGI-1 (semi-private)Reasoning95.5%17 / 126
LiveBench LanguageKnowledge85.5%10 / 66
Artificial Analysis Long Context ReasoningReasoning81.7%33 / 217
OTIS Mock AIME 2024-2025Math97.2% ±1.934 / 200
Artificial Analysis SciCodeCoding57.2%21 / 114
Terminal-Bench 2.1 · Vals AIAgentic77.5% ±0.614 / 75
Terminal-Bench 2.1 · BenchLMAgentic85.8%17 / 86
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal70.4%9 / 39
Critical Physics TasksReasoning14.3%52 / 214
CharXiv ReasoningMultimodal88.7%12 / 49
Mystery Game PuzzlesReasoning37.0% ±4.920 / 80
SWE-benchCoding80.8% ±1.823 / 87
LiveBench MathematicsMath93.5%19 / 66
LiveBench Agentic CodingAgentic58.3%19 / 66
GDPval-AA normalizedAgentic44.6%40 / 133
FrontierMath-Tiers-1-3-v2-PrivateMath71.6% ±2.728 / 90
IOICoding67.8% ±3.614 / 45
LiveBench ReasoningReasoning87.8%26 / 80
MazeBench: visual spatial reasoning in a 3D open worldAgentic1.0%16 / 48
Artificial Analysis Terminal-Bench v2.1Agentic85.8%8 / 21
Vibe Code Bench v1.1Coding70.4% ±4.848 / 123
FrontierMath-Tier-4-v2-PrivateMath36.6% ±7.629 / 72
LiveBench CodingCoding78.9%27 / 66
LiveBench ResearchAgentic76.5%23 / 52
CharXiv Reasoning without toolsMultimodal84.5%8 / 17
OSWorld 2.0Agentic47.9%12 / 25
Code MigrationCoding34.8% ±4.237 / 77
ProofBench v1.1Math58.0% ±5.026 / 54
MirrorCodeCoding41.1% ±12.221 / 39
ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable jobAgentic16.0%10 / 18
DeepSWEAgentic65.3%34 / 59
FrontierCode 1.1 MainCoding43.6%10 / 17
LiveBench Browser UseAgentic56.0%29 / 49
LiveBench Coding 2.0Coding80.5%32 / 52
Terminal-Bench 3.0Agentic14.9%8 / 13
LiveBench DesignAgentic73.2%33 / 52
LiveBench Long AgenticAgentic28.3%33 / 51
Terminal-Bench 4.0Agentic12.1% ±3.042 / 61
Agent Arena task outcomeAgentic−3.538 / 54
LiveBench Data AnalysisReasoning68.0%59 / 80
FrontierSWE v2Coding20.3%17 / 22
Agent Arena steerabilityAgentic−7.342 / 54
Agent Arena command recoveryAgentic−5.440 / 50
Agents' Last ExamAgentic26.3%17 / 21
Furniture AssemblyReasoning26.7% ±5.630 / 37
AutomationBenchAgentic30.4%28 / 32
Terminal-Bench 4.1.0Agentic12.4% ±1.321 / 23
Terminal-Bench 4.0.0Agentic11.2% ±1.323 / 23
LegalBenchLegal87.3% ±0.46 / 149
MedCodeHealthcare53.4% ±2.19 / 106
Finance Agent (v2)Finance59.0% ±0.37 / 79
ARC-AGI-2 (public eval)Reasoning85.3%19 / 122
EMBFinance71.3% ±2.313 / 75
ARC-AGI-1 (public eval)Reasoning96.6%24 / 116
TaxEval v2Finance74.7% ±0.831 / 143
MortgageTaxFinance66.7% ±0.933 / 97
MedScribeHealthcare83.9% ±2.040 / 109
SRE BenchOther4.6% ±1.317 / 36
Terminal-Bench-ScienceScience7.1% ±3.121 / 44
ProgramBenchCoding0.0%34 / 65
Legal Research BenchLegal34.6% ±3.349 / 90
Tax Agent BenchFinance22.8% ±3.048 / 82
CyberBench v1.1Other43.8% ±2.243 / 48
OpenAI MRCR v2 8-needle 64K-128KReasoning97.0%—
Artificial Analysis Harvey LAB-AA v1.0Agentic90.7%—
BioMysteryBench Human SolvableKnowledge87.1%—
LVBenchMultimodal85.4%—
LABBench2: An Improved Benchmark for AI Systems Performing Biology ResearchKnowledge82.1%—
HLE-VerifiedKnowledge53.6%—
BioMysteryBench Human DifficultKnowledge43.5%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYARC Prize FoundationTerms restrict reuseVals AINo licence statedLiveBenchNo licence statedEuroEvalMITTerminal-BenchNo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index70.1 → 68.5
23 Oct 202612 Nov 2026
Output / 1M$3.75 → $3.75
23 Oct 202612 Nov 2026

Questions

How much does Gemini 3.7 Flash cost?
$0.75 per 1M input tokens, $3.75 per 1M output tokens.
How good is Gemini 3.7 Flash?
Its index is 68.5 out of 100. That is place 27 of 283 ranked models.
What is Gemini 3.7 Flash best at?
Knowledge: 71.2 out of 100, place 15 of 268.
How fast is Gemini 3.7 Flash?
85 output tokens per second, 2.4 s to the first token.
What is the context window of Gemini 3.7 Flash?
1.05M tokens.
Is Gemini 3.7 Flash open source?
No. Its weights are not published.
When was Gemini 3.7 Flash released?
August 2026.

Badge

Gemini 3.7 Flash on Inference360: place 27 of 283, index 68.5
HTML
<a href="https://inference360.ai/models/gemini-3-7-flash"><img src="https://inference360.ai/badges/gemini-3-7-flash.svg" alt="Gemini 3.7 Flash on Inference360: place 27 of 283, index 68.5" height="20"></a>
Markdown
[![Gemini 3.7 Flash on Inference360: place 27 of 283, index 68.5](https://inference360.ai/badges/gemini-3-7-flash.svg)](https://inference360.ai/models/gemini-3-7-flash)

Compare with

DeepSeek V4.1 Flash 101768.5Step 5 Preview68.4GPT-5.6 Terra68.9Gemini 3.8 Flash67.8GPT-5.569.1

Lower price

Ling 3.1 Flash67.2 · $0.135/1MDeepSeek V4.1 Flash 101768.5 · $0.40/1MClaude Haiku 5.566.3 · $0.50/1M

More from Google

All models
Gemini 3.8 Flash67.8Gemini 3.8 Flash CyberUnrankedGemini 3.6 Flash63.7Gemini 3.5 Flash-Lite52.0Gemini 3.5 Flash CyberUnranked