Inference
EN
Google

Gemini 3.5 Flash

Gemini 3.5 Flash is a reasoning model from Google. 51 benchmarks count toward its index, in 7 categories.

  • Closed weights
  • Reasoning
  • Input: text · image · video · file · audio
  • Released 2026-05
RankedEnough results for an index.
IndexAll categories, out of 100.64.2±2.8Median 51.0
CoverageIndex weight that has results.95%
SpeedOutput tokens per second.137/sMedian 67/s
First tokenSeconds to the first output token.1.7 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$1.50batch $0.75Median $0.60
Output / 1MUS dollars per 1M output tokens.$9.00batch $4.50 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1483±4 #19Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
google/gemini-3.5-flash
Max outputMaximum output tokens per request.
66K
Knowledge cutoffLast date of training data.
1 Jan 2025
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.15
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$1.50$9.001.05M
Google Vertex AIMaker$1.50$9.001.05M
OpenRouterRouter$1.50$9.001.05M
SAP AI CoreCloud$1.50$9.001.05M

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.14 counted64.747 / 218
CodingCode writing and repair.9 counted60.866 / 209
ReasoningLogic problems and puzzles.9 counted62.257 / 227
MultimodalTasks with images and text.4 counted64.728 / 84
KnowledgeFacts and expert knowledge.6 counted68.527 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.2 counted69.93 / 63
MathMath problems.7 counted61.545 / 168

Results

51 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
APEX-Agents-AAAgentic47.1%1 / 30
Terminal-Bench 2.0 · Vals AIAgentic67.4% ±1.34 / 66
Artificial Analysis MMMU-ProMultimodal84.3%8 / 113
MMLU ProKnowledge89.5% ±0.310 / 135
Massive Multi-discipline Multimodal Understanding ProMultimodal83.6%4 / 53
LiveCodeBenchCoding87.6% ±0.912 / 142
MMMU ProMultimodal88.3% ±0.88 / 92
Artificial Analysis IFBenchInstruction76.3%13 / 146
Artificial Analysis Omniscience AccuracyKnowledge51.9%15 / 164
SWE-Bench verified · Epoch AICoding79.3% ±1.83 / 32
GPQA Diamond · Vals AIKnowledge92.7% ±1.514 / 135
Artificial Analysis GPQA DiamondKnowledge92.2%21 / 199
LiveBench Instruction FollowingInstruction75.6%7 / 66
GPQA diamond · Epoch AIKnowledge92.8% ±1.624 / 217
FrontierMath-2025-02-28-PrivateMath39.0% ±2.99 / 69
τ²-Bench Tool-Agent-User EvaluationAgentic95.3%21 / 149
Graduate-Level Google-Proof Q&AKnowledge92.2%20 / 141
Scientific Code BenchmarkCoding53.1%5 / 35
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal73.2%6 / 39
SWE-bench Verified · SWE-bench/experimentsCoding71.8%10 / 65
SimpleQA VerifiedKnowledge66.2% ±1.514 / 91
Artificial Analysis Humanity's Last ExamKnowledge42.7%36 / 222
MCP AtlasAgentic83.6%9 / 52
ToolathlonAgentic56.5%4 / 22
GPQA Diamond · BenchLMKnowledge92.7%16 / 84
Gert Labs Composite Game BenchmarkAgentic61.9%10 / 52
OTIS Mock AIME 2024-2025Math95.6% ±2.740 / 200
ARC-AGI-2 (semi-private)Reasoning72.1%27 / 131
LiveBench LanguageKnowledge84.6%14 / 66
ARC-AGI-1 (semi-private)Reasoning92.5%28 / 126
FrontierMath-Tier-4-2025-07-01-PrivateMath14.6% ±5.113 / 55
Critical Physics TasksReasoning13.1%55 / 214
Terminal-Bench 2.1 · Vals AIAgentic74.2% ±1.120 / 75
Chess PuzzlesReasoning45.0% ±5.028 / 90
SWE-benchCoding78.8% ±1.828 / 87
Artificial Analysis SciCodeCoding53.9%37 / 114
GDPval-AA normalizedAgentic42.2%45 / 133
Mystery Game PuzzlesReasoning32.0% ±4.729 / 80
OSWorld-VerifiedAgentic78.4%14 / 37
CharXiv ReasoningMultimodal84.2%21 / 49
MRCRv2Reasoning77.3%6 / 14
LiveBench CodingCoding78.2%29 / 66
Humanity's Last ExamKnowledge40.2%35 / 77
FrontierMath-Tiers-1-3-v2-PrivateMath62.8% ±2.941 / 90
Instruction Following BenchmarkInstruction76.3%26 / 57
MazeBench: visual spatial reasoning in a 3D open worldAgentic0.0%23 / 48
Terminal-Bench 2.1 · BenchLMAgentic76.2%43 / 86
SkillsBenchCoding52.7% ±4.518 / 35
Artificial Analysis Long Context ReasoningReasoning69.3%115 / 217
FrontierMath-Tier-4-v2-PrivateMath26.8% ±7.039 / 72
Vibe Code Bench v1.1Coding48.7% ±4.768 / 123
Code MigrationCoding26.7% ±4.145 / 77
LiveBench MathematicsMath88.2%40 / 66
Artificial Analysis AnalystAgentAgentic45.0%17 / 26
LiveBench ReasoningReasoning82.0%53 / 80
SWE-bench ProCoding55.1%55 / 83
LiveBench Agentic CodingAgentic49.0%45 / 66
ProofBench v1.1Math31.0% ±4.638 / 54
Artificial Analysis ITBench-AAAgentic40.3%17 / 24
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal79.8%21 / 26
Terminal-Bench 4.0Agentic6.1% ±1.550 / 61
LiveBench Data AnalysisReasoning64.9%67 / 80
EBR-benchReasoning4.8%29 / 31
cursorBench32Coding48.8%18 / 18
MedCodeHealthcare55.8% ±2.16 / 106
FrontierMath-Tier-4-2025-07-01-PublicMath0.0%3 / 36
FrontierMath-2025-02-28-PublicMath80.0% ±13.36 / 64
Finance Agent (v2)Finance57.9% ±0.213 / 79
MortgageTaxFinance68.1% ±0.919 / 97
ARC-AGI-2 (public eval)Reasoning72.1%27 / 122
ARC-AGI-1 (public eval)Reasoning96.0%28 / 116
TaxEval v2Finance74.4% ±0.836 / 143
CorpFin v2Finance64.7% ±0.934 / 131
LegalBenchLegal83.6% ±0.851 / 149
EMBFinance63.5% ±2.828 / 75
Public Benefits Bench v1Public benefits58.0% ±1.35 / 13
Terminal-Bench-ScienceScience5.7% ±2.822 / 44
ProgramBenchCoding0.0%34 / 65
SWE-bench Multilingual · SWE-bench/experimentsCoding67.0%8 / 14
Legal Research BenchLegal30.8% ±3.253 / 90
Tax Agent BenchFinance21.7% ±3.051 / 82
Public Benefits Bench v1.1Public benefits59.5% ±1.333 / 50
MedScribeHealthcare76.6% ±1.975 / 109
Vals Web Search IndexOther37.0% ±1.911 / 12
OEIS Open LiteMath29.0% ±4.65 / 5
OEIS OpenMath22.2% ±1.93 / 3
Artificial Analysis EnterpriseOps-GymAgentic50.1%—
cursorBench31Coding49.8%—
Blueprint-Bench 2Multimodal33.6%—
MRCR 1MReasoning26.6%—
ResearchClawBenchAgentic18.0%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYVals AINo licence statedARC Prize FoundationTerms restrict reuseLiveBenchNo licence statedSWE-bench teamNo licence statedmodels.devMIT

History

5 releases
Index65.5 → 64.2
23 Oct 202612 Nov 2026
Output / 1M$9.00 → $9.00
23 Oct 202612 Nov 2026

Questions

How much does Gemini 3.5 Flash cost?
$1.50 per 1M input tokens, $9.00 per 1M output tokens.
How good is Gemini 3.5 Flash?
Its index is 64.2 out of 100. That is place 51 of 283 ranked models.
What is Gemini 3.5 Flash best at?
Instruction: 69.9 out of 100, place 3 of 63.
How fast is Gemini 3.5 Flash?
137 output tokens per second, 1.7 s to the first token.
What is the context window of Gemini 3.5 Flash?
1.05M tokens.
Is Gemini 3.5 Flash open source?
No. Its weights are not published.
When was Gemini 3.5 Flash released?
May 2026.

Badge

Gemini 3.5 Flash on Inference360: place 51 of 283, index 64.2
HTML
<a href="https://inference360.ai/models/gemini-3-5-flash"><img src="https://inference360.ai/badges/gemini-3-5-flash.svg" alt="Gemini 3.5 Flash on Inference360: place 51 of 283, index 64.2" height="20"></a>
Markdown
[![Gemini 3.5 Flash on Inference360: place 51 of 283, index 64.2](https://inference360.ai/badges/gemini-3-5-flash.svg)](https://inference360.ai/models/gemini-3-5-flash)

Compare with

Claude Sonnet 564.2Qwen3.8 Max Preview64.2MiMo-V2.6-Flash64.3Sakana Fugu63.9dots3-note Preview63.4

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
Gemma 4 12B39.0Gemini Pro LatestUnrankedGemini Flash LatestUnrankedNano Banana Pro (Gemini 3 Pro Image)UnrankedNano Banana 2 (Gemini 3.1 Flash Image)Unranked