Inference
EN
Google

Gemini 3 Pro

Gemini 3 Pro is a standard model from Google. 35 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Standard
  • Input: text
  • Released 2025-11
RankedEnough results for an index.
IndexAll categories, out of 100.57.9±2.7Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.109/sMedian 67/s
First tokenSeconds to the first output token.32.6 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$2.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$12.00Median $2.55
ContextMaximum tokens in one request.2MMedian 262K
EloLMArena rating and rank.1480±4 #21Median 1422

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.5 counted56.294 / 218
CodingCode writing and repair.4 counted56.488 / 209
ReasoningLogic problems and puzzles.4 counted57.275 / 227
MultimodalTasks with images and text.11 counted54.855 / 84
KnowledgeFacts and expert knowledge.4 counted66.139 / 268
MultilingualTasks in many languages.1 counted60.1—
InstructionTasks with strict prompt rules.1 counted54.8—
MathMath problems.5 counted58.252 / 168

Results

35 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
MATH 500Math96.4% ±0.81 / 59
Massive Multi-discipline Multimodal UnderstandingMultimodal87.2%1 / 28
EuroEval PolishMultilingual64.8% ±0.58 / 209
AIMEMath96.7% ±0.44 / 95
SWE-bench Verified · swebench.comCoding74.2%2 / 46
MMLU ProKnowledge90.1% ±0.36 / 135
Artificial Analysis MMLU-ProKnowledge89.8%1 / 15
Artificial Analysis Omniscience AccuracyKnowledgeAbove fitFar above this model's index.55.8%11 / 164
Artificial Analysis LiveCodeBenchCoding91.7%1 / 12
GPQA diamond · Epoch AIKnowledge92.6% ±1.727 / 217
GPQA Diamond · Vals AIKnowledge91.7% ±1.418 / 135
Gert Labs Composite Game BenchmarkAgentic63.2%7 / 52
Artificial Analysis GPQA DiamondKnowledge90.8%30 / 199
MMMU ProMultimodal87.5% ±0.814 / 92
LiveCodeBenchCoding86.4% ±1.023 / 142
FrontierMath-2025-02-28-PrivateMath37.6% ±2.812 / 69
FrontierMath-Tier-4-2025-07-01-PrivateMath18.8% ±5.710 / 55
Terminal-Bench 1.0Agentic51.3% ±5.69 / 47
Artificial Analysis MMMU-ProMultimodal80.2%22 / 113
CountBenchMultimodal97.3%2 / 10
Artificial Analysis Humanity's Last ExamKnowledge39.7%48 / 222
Massive Multi-discipline Multimodal Understanding ProMultimodal81.0%12 / 53
Terminal-Bench 2.0 · Vals AIAgentic55.1% ±5.315 / 66
IOI v1Coding38.8% ±11.514 / 61
SWE-bench Verified · SWE-bench/experimentsCoding69.6%17 / 65
τ²-bench AirlineAgentic80.5%6 / 22
SimpleVQAMultimodal73.2%5 / 18
OTIS Mock AIME 2024-2025Math91.4% ±3.756 / 200
Artificial Analysis IFBenchInstruction70.4%42 / 146
τ²-Bench Tool-Agent-User EvaluationAgentic87.1%48 / 149
Critical Physics TasksReasoning9.1%70 / 214
ERQAMultimodal70.5%6 / 18
Artificial Analysis Long Context ReasoningReasoning76.0%88 / 217
SWE-benchCoding76.4% ±1.937 / 87
ARC-AGI-2 (semi-private)Reasoning31.1%56 / 131
ARC-AGI-1 (semi-private)Reasoning75.0%56 / 126
τ²-bench TelecomAgentic91.0%9 / 18
RealWorldQAMultimodal83.3%9 / 18
ScreenSpot ProMultimodal72.7%11 / 20
CharXiv ReasoningMultimodal81.4%28 / 49
MathVisionMultimodal86.6%14 / 24
V*Multimodal88.0%9 / 14
τ²-bench RetailAgentic75.9%15 / 23
SWE-Bench verified · Epoch AICoding72.9% ±2.021 / 32
τ²-bench BankingAgentic18.0%22 / 28
Vibe Code Bench v1.1Coding14.3% ±3.199 / 123
RefCOCO averageMultimodal84.1%9 / 10
JobBenchAgentic11.4%33 / 36
LegalBenchLegal87.0% ±0.47 / 149
MortgageTaxFinance69.1% ±0.97 / 97
FrontierMath-Tier-4-2025-07-01-PublicMath0.0%3 / 36
MedQAHealthcare96.0% ±0.28 / 95
FrontierMath-2025-02-28-PublicMath80.0% ±13.36 / 64
MGSMMultilingual93.9% ±1.47 / 74
MedCodeHealthcare52.2% ±2.114 / 106
Finance Agent v1.1Finance55.2% ±2.814 / 51
CorpFin v2Finance63.7% ±0.940 / 131
SWE-bench Multilingual · SWE-bench/experimentsCoding68.7%5 / 14
SWE-bench Multilingual · swebench.comMultilingual68.7%5 / 12
TaxEval v2Finance72.6% ±0.960 / 143
CaseLaw v2Legal53.1% ±0.241 / 53
MedScribeHealthcare72.0% ±1.991 / 109
Artificial Analysis AIME 2025Math95.7%—
AI2D test splitMultimodal94.1%—
Artificial Analysis Global-MMLU-LiteMultilingual92.2%—
OmniDocBench 1.5Multimodal88.5%—
Video-MME with subtitleMultimodal88.4%—
Video-MME without subtitleMultimodal87.7%—
VideoMMMUMultimodal87.6%—
We-MathMultimodal86.9%—
DynaMathMultimodal85.1%—
MLVU mean averageMultimodal83.0%—
CC-OCRMultimodal79.0%—
MMLongBench-DocMultimodal60.5%—
ODINW13Multimodal46.3%—

Sources

BenchLMCC BY-NC 4.0Vals AINo licence statedEpoch AICC BYSierra ResearchMITARC Prize FoundationTerms restrict reuseSWE-bench teamNo licence statedEuroEvalMIT

History

5 releases
Index59.6 → 57.9
23 Oct 202612 Nov 2026
Output / 1M$12.00 → $12.00
23 Oct 202612 Nov 2026

Questions

How much does Gemini 3 Pro cost?
$2.00 per 1M input tokens, $12.00 per 1M output tokens.
How good is Gemini 3 Pro?
Its index is 57.9 out of 100. That is place 85 of 283 ranked models.
What is Gemini 3 Pro best at?
Knowledge: 66.1 out of 100, place 39 of 268.
How fast is Gemini 3 Pro?
109 output tokens per second, 32.6 s to the first token.
What is the context window of Gemini 3 Pro?
2M tokens.
Is Gemini 3 Pro open source?
No. Its weights are not published.
When was Gemini 3 Pro released?
November 2025.

Badge

Gemini 3 Pro on Inference360: place 85 of 283, index 57.9
HTML
<a href="https://inference360.ai/models/gemini-3-pro"><img src="https://inference360.ai/badges/gemini-3-pro.svg" alt="Gemini 3 Pro on Inference360: place 85 of 283, index 57.9" height="20"></a>
Markdown
[![Gemini 3 Pro on Inference360: place 85 of 283, index 57.9](https://inference360.ai/badges/gemini-3-pro.svg)](https://inference360.ai/models/gemini-3-pro)

Compare with

Claude Opus 4.658.0Kimi K2.7 Code57.7Qwen3.7 Plus57.7Kimi K2.658.2Gemini 3 Flash55.0

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeGLM-5.3-Flash63.1 · $0.078/1M

More from Google

All models
Nano Banana Pro (Gemini 3 Pro Image Preview)UnrankedGemini 3 Flash55.0Gemini 2.5 Flash Native Audio Preview (12-2025)UnrankedGemini 3 Flash Preview57.5Nano Banana (Gemini 2.5 Flash Image)Unranked