The LLM benchmark aggregator.

Gemini 3.1 Flash-Lite

Google

Strongest in Factual grounding (#40 of 104), weakest in Agents (#261 of 298). Above par in 10 of 20 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 4.8 and Claude Fable 5 and ahead of GLM-4.6 and GPT OSS 120B.

§ 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.

Core abilities54.9−10.4#45/1991/4

General intelligence57.1−12.3#41/1781/3

Reasoning53.4−12#60/1951/5

Mathematics54.5−9.3#61/1521/2

Science53.5−10#65/1221/1

Human preference59.8−9#91/3541/1

Knowledge49−13.5#127/1972/3

Academic knowledge56.3−7.6#44/1231/1

Document parsing47.8−15.5#54/831/1

Professional46.3−16.1#142/1941/2

Finance48.5−13.9#108/1621/1

Medical46.1−20.4#110/1491/1

Legal42.8−23.1#140/1731/1

Safety51.4−10#144/3291/3

Factual grounding54.4−16.3#40/1041/1

Coding41.7−24.5#158/1791/6

Agentic coding41.1−25.8#155/1731/5

Agents42.1−24.3#261/2982/6

Knowledge work37.3−33.4#166/2082/2

§ 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 3.1 Flash-Lite, left for the other.

GPT-4o miniwon 15 · lost 2

Llama 4 Scout Instructwon 15 · lost 2

Mistral Medium 3.5won 15 · lost 2

GPT-5 nanowon 14 · lost 2

NVIDIA Nemotron 3.5 Lightning 30B A3Bwon 12 · lost 2

Magistral Small 1.2won 10 · lost 2

Mistral Medium 3.1won 11 · lost 3

GPT-4.1 nanowon 14 · lost 4

Command Awon 14 · lost 4

Mistral Small 3.1 24B Instructwon 10 · lost 3

Gemini 1.5 Prowon 9 · lost 3

GPT OSS 20Bwon 12 · lost 4

Claude Haiku 4.5won 13 · lost 5

Llama 3.3 70B Instructwon 10 · lost 4

Mistral Large 2.1won 9 · lost 4

Gemini 2.0 Flashwon 11 · lost 5

Gemini 2.5 Flash-Litewon 11 · lost 5

Llama 4 Maverick Instructwon 12 · lost 6

Mistral Large 3won 12 · lost 7

Kimi K2won 11 · lost 7

Claude 3.5 Haikuwon 9 · lost 6

GPT-5.4 nanowon 12 · lost 8

GPT-4owon 10 · lost 7

Grok 4 Fastwon 10 · lost 7

DeepSeek V3won 11 · lost 8

GLM-4.5won 8 · lost 6

O3 Miniwon 9 · lost 7

DeepSeek R1won 10 · lost 8

NVIDIA Nemotron 3 Ultra 550B A55Bwon 10 · lost 8

Grok 3won 9 · lost 8

Grok 4.3won 9 · lost 8

GPT OSS 120Bwon 10 · lost 9

GLM-4.6won 10 · lost 9

Grok 4.1 Fastwon 8 · lost 8

Gemma 4 31Bwon 8 · lost 8

GPT-4.1 miniwon 9 · lost 9

Gemini 2.5 Flashwon 9 · lost 9

DeepSeek V3.2won 9 · lost 10

MiniMax M2.5won 8 · lost 9

GLM-4.7won 8 · lost 9

MiniMax M2.1won 7 · lost 8

Mistral Large 4won 7 · lost 8

MiMo V2.5won 7 · lost 8

Claude 3.5 Sonnetwon 6 · lost 7

GPT-4.1won 8 · lost 10

Qwen3 235B A22B Instructwon 8 · lost 10

Qwen3.8 27Bwon 8 · lost 10

Grok 3 Miniwon 7 · lost 9

O1won 6 · lost 8

Claude Sonnet 4won 8 · lost 11

Inkling Smallwon 7 · lost 10

MiniMax M2.7won 7 · lost 10

Claude Opus 4won 7 · lost 10

Claude 3.7 Sonnetwon 6 · lost 9

Qwen3.5 Flashwon 6 · lost 9

GPT-5 miniwon 8 · lost 12

O4 Miniwon 7 · lost 11

Ling 3.0 Flashwon 5 · lost 8

Claude Opus 4.1won 5 · lost 8

Kimi K2 Thinkingwon 5 · lost 9

Qwen3 Maxwon 5 · lost 10

MiMo V2.6 Flashwon 5 · lost 10

Grok 4.20won 5 · lost 10

MiniMax M3won 6 · lost 12

DeepSeek V4 Flashwon 6 · lost 12

GPT-5.4 miniwon 6 · lost 12

GLM-5.3 Flashwon 6 · lost 13

Step 5won 4 · lost 9

Grok 4won 5 · lost 12

Inklingwon 5 · lost 13

Gemini 3.5 Flash-Litewon 5 · lost 13

Kimi K2.6won 5 · lost 13

GLM-5.2won 5 · lost 13

MiMo V2.6 Prowon 4 · lost 11

GPT-5.6 Lunawon 5 · lost 14

DeepSeek V4 Prowon 5 · lost 14

Claude Haiku 5.5won 4 · lost 12

Qwen3.6 Pluswon 4 · lost 12

MiMo V2.5 Prowon 4 · lost 12

O3won 4 · lost 13

GPT-6 Lunawon 4 · lost 13

GLM-5won 4 · lost 13

Qwen3.7 Maxwon 4 · lost 13

GLM-5.3won 4 · lost 13

Gemini 2.5 Prowon 4 · lost 15

Kimi K2.5won 4 · lost 15

Claude Sonnet 4.5won 4 · lost 15

Grok 4.7won 3 · lost 12

Muse Spark 1.2won 3 · lost 12

Muse Spark 1.1won 3 · lost 14

GPT-5.2won 3 · lost 15

Grok 4.6won 3 · lost 15

Claude Sonnet 4.6won 3 · lost 15

Claude Opus 4.5won 3 · lost 15

DeepSeek V4.1 Flashwon 3 · lost 15

Gemini 3.6 Flashwon 3 · lost 15

Qwen3.8 Maxwon 3 · lost 15

Gemini 3.1 Prowon 3 · lost 16

Gemini 3.8 Flashwon 3 · lost 16

Qwen 3.5 Pluswon 2 · lost 11

Gemini 3 Flashwon 2 · lost 14

Gemini 4 Argonwon 2 · lost 14

GPT-5.1won 2 · lost 15

Kimi K3won 2 · lost 15

Grok 4.5won 2 · lost 16

GPT-6 Astrawon 2 · lost 16

Gemini 3.5 Flashwon 2 · lost 16

Gemini 3 Prowon 2 · lost 16

GPT-5won 2 · lost 17

GPT-5.6 Terrawon 2 · lost 17

GPT-5.4won 2 · lost 18

GPT-5.6 Solwon 2 · lost 18

Claude Opus 4.7won 2 · lost 18

Muse Spark 1.3won 1 · lost 12

Claude Sonnet 5.5won 1 · lost 15

GPT-6 Solwon 1 · lost 16

Gemini 3.7 Flashwon 1 · lost 16

Claude Opus 5won 1 · lost 16

Claude Sonnet 5won 1 · lost 17

Claude Opus 4.6won 1 · lost 17

Claude Opus 5.5won 1 · lost 17

GPT-5.5won 1 · lost 19

Muse Wonder Fullwon 0 · lost 12

Kimi K3.1 Prowon 0 · lost 12

Muse Wonderwon 0 · lost 13

Muse Sparkwon 0 · lost 15

GLM-5.1won 0 · lost 16

GPT-6.1 Solwon 0 · lost 16

Claude Fable 5.1won 0 · lost 18

Claude Fable 5won 0 · lost 18

Claude Opus 4.8won 0 · lost 19

§ 3 · Sources

Where the numbers come from

7 publications, 25 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-10

LMArena Text 1433

GDPval-AA ↗1 measureread 2026-11-10

GDPval-AA 438

AA-Briefcase ↗1 measureread 2026-11-10

AA-Briefcase 208

Kagi LLM Benchmark ↗1 measureread 2026-11-10

Kagi LLM Benchmark 67.2%

Vals.ai ↗14 measuresread 2026-11-10

Vals · Legal Research Bench 3.37%Vals · CaseLaw 54.98%Vals · LegalBench 83.76%Vals · Finance Agent 29.99%Vals · CorpFin 59.36%Vals · TaxEval 71.79%Vals · MortgageTax 68.04%Vals · MedCode 47.6%Vals · MedScribe 63.9%Vals · SWE-bench Verified 62.8% (Mini-SWE-agent)Vals · Code Migration 4.61%Vals · GPQA Diamond 81.06%Vals · MMLU Pro 86.24%Vals · AIME 83.33%

ParseBench ↗6 measuresread 2026-11-10

ParseBench · Overall 55.84%ParseBench · Tables 76.8%ParseBench · Charts 30.89%ParseBench · Content faithfulness 82.84%ParseBench · Semantic formatting 56.61%ParseBench · Visual grounding 32.08%

Vectara Hallucination Leaderboard ↗1 measureread 2026-11-10

Vectara · Factual consistency 91.8%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=gemini-3-1-flash-lite)](https://publicai.io/model-index/m/gemini-3-1-flash-lite)