GoogleReasoning modelAuto-detected

Gemini 3.1 Flash Lite

Name: Gemini 3.1 Flash Lite benchmark results
Creator: BenchLeader

Best configuration ranks #202 of 610 on the BenchLeader Index at 54.3 ±4.2. Last measured 8 Sept 2026. Released 3 Mar 2026.

Blended price: $0.563/M; $0.250 in · $1.50 out
Output speed: 303 tok/s
First answer: 5.78 s; first token 0.62 s
Context: 1.0M

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index54
Reasoning50
Coding48
Agents & tools46
Knowledge54
Instruction following62
Human preference62
Multimodal59
Long context64
Composite49

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
minimal	–	–	303 tok/s	5.78 s	$0.0006	Maths 40 · Reasoning 54
low	–	–	303 tok/s	5.78 s	$0.0006	Maths 41 · Reasoning 55
high	49.1	#310	303 tok/s	5.78 s	$0.0006	Agents & tools 30 · Knowledge 55 · Maths 51 · Reasoning 60
defaultbest	54.3	#202	303 tok/s	5.78 s	$0.0006	Agents & tools 46 · Coding 48 · Composite 49 · Human preference 62 · Instruction following 62 · Knowledge 54 · Long context 64 · Multimodal 59 · Reasoning 50

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	minimal	low	high	default	Source
GPQA Diamond	73.7%#144	74.2%#141	81.8%#105	–	Epoch AI Benchmarking Hub
Humanity's Last Exam	–	–	–	8.6%#27	Scale AI / CAIS
LMArena Hard Prompts	–	–	–	1445#102	LMArena
GPQA Diamond (AA)not in index	–	–	–	82.2%#169	Artificial Analysis
Humanity's Last Exam (AA)not in index	–	–	–	17.2%#178	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	81.1%#63	–	Vals AI
Kagi LLM Benchmark	–	–	–	67.2%#34	Kagi LLM Benchmark

Coding

Benchmark	minimal	low	high	default	Source
SciCode	–	–	–	41.9%#100	SciCode
WeirdML	–	–	–	52.2%#65	WeirdML
LMArena Coding	–	–	–	1457#120	LMArena
LMArena WebDev	–	–	–	1254#105	LMArena
SciCode (AA)not in index	–	–	–	43.4%#99	Artificial Analysis
LiveCodeBench	–	–	–	80.1%#68	Vals AI
SWE-bench (Vals)not in index	–	–	62.8%#72	–	Vals AI

Agents & tools

Benchmark	minimal	low	high	default	Source
APEX-Agents	–	–	–	13.0%#49	Mercor
Terminal-Bench Hard	–	–	–	24.2%#137	Artificial Analysis
τ²-Bench Telecom (AA)not in index	–	–	–	31.3%#247	Artificial Analysis
Terminal-Bench 2.1 (Vals)	–	–	34.1%#58	–	Vals AI
MCP Atlas	–	–	57.1%#26	–	Scale AI SEAL

Maths

Benchmark	minimal	low	high	default	Source
FrontierMath Tiers 1–3	21.4%#71	22.5%#69	27.7%#62	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	37.8%#182	44.4%#175	80.0%#99	–	Epoch AI Benchmarking Hub
AIME (Vals)	–	–	83.3%#43	–	Vals AI

Knowledge

Benchmark	minimal	low	high	default	Source
AA-Omniscience	–	–	–	-16.4#168	Artificial Analysis
MMLU-Pro	–	–	–	86.2%#43	Vals AI
LegalBench	–	–	83.8%#44	–	Vals AI
CorpFin	–	–	59.4%#76	–	Vals AI
TaxEval	–	–	71.8%#71	–	Vals AI
MultiNRC	–	–	–	25.0%#29	Scale AI SEAL

Instruction following

Benchmark	minimal	low	high	default	Source
IFBench	–	–	–	77.2%#17	Artificial Analysis
MultiChallenge	–	–	–	60.6%#10	Scale AI SEAL
TutorBench	–	–	–	51.5%#15	Scale AI SEAL

Human preference

Benchmark	minimal	low	high	default	Source	Trend
LMArena Text	–	–	–	1433#90	LMArena

Multimodal

Benchmark	minimal	low	high	default	Source
LMArena Vision	–	–	–	1241#57	LMArena
MMMU-Pro	–	–	–	75.5%#69	Artificial Analysis
VISTA	–	–	–	46.9%#19	Scale AI SEAL

Long context

Benchmark	minimal	low	high	default	Source	Trend
AA-LCR	–	–	–	74.3%#116	Artificial Analysis

Composite

Benchmark	minimal	low	high	default	Source
Epoch Capabilities Indexnot in index	–	–	–	144.5#74	Epoch AI Benchmarking Hub
AA Intelligence Index	–	–	–	16.0#212	Artificial Analysis
Vals Indexnot in index	–	–	15.5#51	–	Vals AI

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Google AI Studio Flex	146 tok/s	0.49 s	$0.125	$0.750	1.0M	–
Google Vertex Priority	143 tok/s	0.66 s	$0.450	$2.70	1.0M	–
Google AI Studio	94 tok/s	0.59 s	$0.250	$1.50	1.0M	–
Google Vertex	71 tok/s	0.83 s	$0.250	$1.50	1.0M	–
Google AI Studio Priority	51 tok/s	0.55 s	$0.450	$2.70	1.0M	–
Google Vertex Flex	6 tok/s	17 s	$0.125	$0.750	1.0M	–

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.025 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0006	$0.0005	6.8 s
Summarise a 30-page report	12,000 / 600	$0.0039	$0.0019	7.8 s
Code edit	6,000 / 1,500	$0.0038	$0.0027	10.7 s
Agentic coding session	60,000 / 4,000	$0.021	$0.011	19.0 s
Structured extraction	2,000 / 200	$0.0008	$0.0005	6.4 s