Request a Pilot
‹ PublicAI Index

The LLM benchmark aggregator.

Gemini 3 Flash

Google

Strongest in Knowledge (#10 of 197, on 2 of its 3 boards), weakest in Safety (#244 of 329). Above par in 12 of 17 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Fable 5.1 and ahead of GLM-5.2 and Kimi K2.6.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Knowledge59.5#10/197
Document parsing59.9#13/83
Academic knowledge59.2#19/123
Human preference63.7#37/354
Human preference63.7#37/354
Reasoning55.4#40/195
Mathematics57.8#25/152
Science58.2#37/122
Reasoning52#56/147
Professional52.9#63/194
Medical56#35/149
Finance54.9#50/162
Legal48.2#104/173
Coding43.3#140/179
Agentic coding42.9#136/173
Safety47.2#244/329
Factual grounding41#88/104

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Gemini 3 Flash, left for the other.

Grok 2won 11 · lost 0
Mistral Medium 3.1won 12 · lost 0
Mistral Medium 3.5won 13 · lost 0
Qwen3.6 Pluswon 13 · lost 0
GPT OSS 120Bwon 15 · lost 1
Grok 4 Fastwon 15 · lost 1
Mistral Large 3won 15 · lost 1
Grok 4.1 Fastwon 15 · lost 1
GPT-4.1 nanowon 14 · lost 1
GPT-4o miniwon 13 · lost 1
Gemini 1.5 Prowon 12 · lost 1
Command R+won 11 · lost 1
GPT OSS 20Bwon 11 · lost 1
GLM-4.5won 11 · lost 1
Ling 3.0 Flashwon 10 · lost 1
Gemini 1.5 Flashwon 10 · lost 1
Claude Haiku 4.5won 15 · lost 2
Gemini 3.1 Flash-Litewon 14 · lost 2
Kimi K2.5won 14 · lost 2
Command Awon 13 · lost 2
GLM-4.6won 13 · lost 2
Inklingwon 13 · lost 2
Gemini 3.5 Flash-Litewon 13 · lost 2
Llama 4 Scout Instructwon 12 · lost 2
GPT-5 nanowon 12 · lost 2
Inkling Smallwon 12 · lost 2
Grok 4.3won 12 · lost 2
Qwen3 Maxwon 11 · lost 2
MiMo V2.5won 11 · lost 2
Gemma 4 31Bwon 11 · lost 2
Mistral Large 2.1won 10 · lost 2
MiniMax M2.7won 12 · lost 3
DeepSeek R1won 12 · lost 3
GLM-5.1won 12 · lost 3
Claude 3.5 Sonnetwon 11 · lost 3
GPT-6 Lunawon 11 · lost 3
MiMo V2.5 Prowon 11 · lost 3
Llama 3.3 70B Instructwon 10 · lost 3
Claude 3.5 Haikuwon 10 · lost 3
Gemini 2.5 Flash-Litewon 10 · lost 3
GPT-5.4 nanowon 13 · lost 4
GPT-4owon 12 · lost 4
MiniMax M2.1won 12 · lost 4
MiniMax M2.5won 12 · lost 4
O4 Miniwon 12 · lost 4
Grok 3won 12 · lost 4
Kimi K2won 12 · lost 4
DeepSeek V3.2won 12 · lost 4
Gemini 2.0 Flashwon 11 · lost 4
O3 Miniwon 11 · lost 4
GPT-4.1 miniwon 11 · lost 4
Grok 4won 11 · lost 4
Qwen3.8 27Bwon 11 · lost 4
GPT-5.4 miniwon 11 · lost 4
Claude Opus 4.1won 10 · lost 4
Grok 4.20won 10 · lost 4
GPT-5 miniwon 12 · lost 5
DeepSeek V3won 11 · lost 5
Claude Sonnet 4won 11 · lost 5
Gemini 2.5 Flashwon 11 · lost 5
Qwen3 235B A22B Instructwon 11 · lost 5
Claude Opus 4won 11 · lost 5
GLM-4.7won 11 · lost 5
GLM-5won 11 · lost 5
Mistral Large 4won 8 · lost 4
Grok 3 Miniwon 10 · lost 5
Claude 3.7 Sonnetwon 10 · lost 5
O3won 10 · lost 5
DeepSeek V4 Flashwon 10 · lost 5
DeepSeek V4won 7 · lost 4
MiMo V2.6 Flashwon 7 · lost 4
GPT-5won 10 · lost 6
GPT-5.6 Lunawon 10 · lost 6
Claude Sonnet 4.5won 10 · lost 6
Kimi K2 Thinkingwon 8 · lost 5
Claude Haiku 5.5won 8 · lost 5
Qwen3.5 Flashwon 9 · lost 6
O1won 9 · lost 6
GPT-4.1won 9 · lost 7
Gemini 2.5 Prowon 9 · lost 7
GPT-5.1won 9 · lost 7
DeepSeek V4 Prowon 9 · lost 7
Kimi K2.6won 8 · lost 7
GLM-5.2won 8 · lost 7
Grok 4.7won 6 · lost 6
Claude Sonnet 4.6won 8 · lost 8
MiniMax M3won 7 · lost 8
Grok 4.5won 7 · lost 8
Qwen 3.5 Pluswon 6 · lost 7
Claude Sonnet 5won 7 · lost 9
GPT-5.6 Terrawon 7 · lost 9
GLM-5.3 Flashwon 7 · lost 9
GPT-6 Solwon 6 · lost 9
Qwen3.7 Maxwon 6 · lost 9
Gemini 3.6 Flashwon 6 · lost 10
GLM-5.3won 5 · lost 9
Grok 4.6won 5 · lost 10
GPT-5.2won 5 · lost 11
Claude Opus 4.5won 5 · lost 11
Gemini 3.5 Flashwon 5 · lost 11
Claude Sonnet 5.5won 4 · lost 9
GPT-6.1 Solwon 4 · lost 9
Muse Spark 1.2won 4 · lost 9
DeepSeek V4.1 Flashwon 4 · lost 11
Qwen3.8 Maxwon 4 · lost 11
MiMo V2.6 Prowon 3 · lost 9
Gemini 3.8 Flashwon 4 · lost 12
GPT-5.4won 4 · lost 13
Gemini 3.1 Prowon 4 · lost 13
Kimi K3won 3 · lost 11
Muse Spark 1.1won 3 · lost 11
GPT-6 Astrawon 3 · lost 12
Gemini 3.7 Flashwon 3 · lost 12
Claude Opus 4.8won 3 · lost 13
Gemini 3 Prowon 3 · lost 13
GPT-5.5won 3 · lost 14
Claude Opus 4.6won 3 · lost 14
Muse Sparkwon 2 · lost 11
Claude Fable 5won 2 · lost 13
Claude Opus 4.7won 2 · lost 15
Claude Opus 5won 1 · lost 14
GPT-5.6 Solwon 1 · lost 16
Gemini 4 Argonwon 0 · lost 13
Claude Fable 5.1won 0 · lost 15
Claude Opus 5.5won 0 · lost 15

§ 3 · Sources

Where the numbers come from

6 publications, 26 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-10
LMArena Text 1473
ARC-AGI-2 ↗1 measureread 2026-11-10
ARC-AGI-2 33.6%
SimpleBench ↗1 measureread 2026-11-10
SimpleBench 61.1%
Vals.ai ↗16 measuresread 2026-11-10
Vals · Legal Research Bench 18.27%Vals · CaseLaw 55.84%Vals · LegalBench 86.86%Vals · Finance Agent 42.55%Vals · CorpFin 66.43%Vals · TaxEval 73.88%Vals · MortgageTax 68.72%Vals · MedQA 95.81%Vals · MedCode 55.92%Vals · MedScribe 69.92%Vals · SWE-bench Verified 75% (Mini-SWE-agent)Vals · Vibe Code Bench 20.2% (OpenHands)Vals · Code Migration 6.39%Vals · GPQA Diamond 87.88%Vals · MMLU Pro 88.59%Vals · AIME 95.63%
ParseBench ↗6 measuresread 2026-11-10
ParseBench · Overall 73.98%ParseBench · Tables 91.19%ParseBench · Charts 64.05%ParseBench · Content faithfulness 89.74%ParseBench · Semantic formatting 62.56%ParseBench · Visual grounding 62.37%
Vectara · Factual consistency 86.5%

Badge

PublicAI Index badge for Gemini 3 Flash[![PublicAI Index](https://publicai.io/model-index/badge?model=gemini-3-flash)](https://publicai.io/model-index/m/gemini-3-flash)