Model Profile

gemini-3-pro-preview

Name: gemini-3-pro-preview
Rating: 5.0 (293 reviews)
Author: google

External Benchmark Shadowexternal_benchmark_shadowpublic

4,096 ctx

Use this page to decide where this model is a strong fit. Rankings below are benchmark-backed by use case, with explicit confidence and contributor metrics.

Identity

ID: external/google/gemini-3-pro-preview

Author: google

Origin: external_benchmark_shadow

Arch: unknown

Benchmark Coverage

Scored use cases: 12

Avg confidence: 55.7%

Evidence points: 293

Raw rows: 714

Weighted rows: 63

Catalog Metadata

Parameters: unknown

Context window: 4096

Downloads: 0

Price / 1M tokens: $4.50 (blended 3:1)

Intelligence Profile

Dimension Breakdown

IQ28 benchmarks

82.2%

EQ9 benchmarks

88.7%

Accuracy10 benchmarks

88.6%

Creativity2 benchmarks

79.5%*

Based1 benchmark

29.0%*

* Low confidence — limited benchmark evidence for this dimension

5/5 dimensions scored · Last updated Apr 21, 2026

Benchmark Signals

Click through to the benchmark source behind this model profile.

BFCL Memory Official

Memory Acc

11.4%

Normalized value 98.1% · confidence 100.0%

Strongest impact in Adult ERP roleplay (explicit)

bfcl_memory_official.memory_acc · Apr 1, 2026

BFCL Multi-turn Official

Multi Turn Acc

7.9%

Normalized value 81.6% · confidence 100.0%

Strongest impact in Casual chat companion

bfcl_multiturn_official.multi_turn_acc · Apr 1, 2026

BFCL Relevance Detection Official

Relevance Detection

5.6%

Normalized value 75.0% · confidence 100.0%

Strongest impact in Casual chat companion

bfcl_relevance_detection_official.relevance_detection · Apr 1, 2026

BasedAGI Log Triage Eval

overall_score_pct

4.2%

Normalized value 86.7% · confidence 100.0%

Strongest impact in Log triage

basedagi_log_triage_eval.overall_score_pct · Apr 1, 2026

Vals Finance Agent

overall_accuracy_pct

3.5%

Normalized value 87.0% · confidence 100.0%

Strongest impact in Thesis red teaming

vals_finance_agent.overall_accuracy_pct · Mar 31, 2026

UGI Leaderboard

Writing ✍️

3.4%

Normalized value 93.9% · confidence 100.0%

Strongest impact in Adult ERP roleplay (explicit)

ugi_main.writing · Apr 1, 2026

Coverage Diagnostics

actively scored

Use-Case Scores

151

Total Measurements

714

Weighted Measurements

Weighted Sources

Raw Source Coverage

vals_mmlu_pro 60ugi_main 57vals_mgsm 48vals_finance_agent 40vals_multimodal_index 32bfcl_adjacent_public 30

Weighted Source Coverage

vectara_hhem_leaderboard 12vals_finance_agent 5basedagi_doc_summarization_eval 4basedagi_kb_qna_eval 4basedagi_log_triage_eval 4basedagi_support_bot_eval 4

Best Use Cases for This Model

Use Case	Vertical	Score	Confidence	Evidence	Top Contributor
Casual chat companion use_case.companion.casual_chat	companion	50.1%	60.5%	24	BFCL Memory Official: Memory Acc
Tarot-style reading use_case.spiritual.tarot_reading	companion	50.1%	60.5%	24	BFCL Memory Official: Memory Acc
Life coaching and goal planning use_case.companion.life_coaching	companion	50.1%	60.5%	24	BFCL Memory Official: Memory Acc
Adult ERP roleplay (explicit) use_case.adult.erp_roleplay	adult	49.3%	59.2%	24	BFCL Memory Official: Memory Acc
Empathetic support chat use_case.companion.empathy_support_chat	companion	48.7%	58.8%	24	BFCL Memory Official: Memory Acc
Mindfulness and meditation scripts use_case.wellness.mindfulness_scripts	companion	48.5%	58.5%	24	BFCL Memory Official: Memory Acc
SFW roleplay and simulation use_case.creative.sfw_roleplay_simulation	creative	45.9%	55.2%	24	BFCL Memory Official: Memory Acc
NPC dialogue use_case.gaming.npc_dialogue	creative	44.2%	53.2%	24	BFCL Memory Official: Memory Acc
Interactive fiction / DM use_case.creative.interactive_fiction_dm	creative	44.2%	53.2%	24	BFCL Memory Official: Memory Acc
Adult erotica (long-form, explicit) use_case.adult.erotica_longform	adult	43.6%	52.4%	24	BFCL Memory Official: Memory Acc
Thesis red teaming use_case.fin.thesis_red_team	finance	37.6%	50.1%	27	Vals Finance Agent: overall_accuracy_pct
Log triage use_case.sre.log_triage	devops_sre	35.6%	45.8%	26	BasedAGI Log Triage Eval: overall_score_pct