GoogleReasoning modelCompare with another model

Gemini 3.7 Flash

Reasoning effort

Gemini 3.7 Flash is a Google proprietary reasoning model, released 13 Aug 2026. Its best configuration (high reasoning effort) ranks #61 of 775 on the BenchLeader Index at 63.5 ±3.5, in the upper half. The ± is the point: 147 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (73) and lowest in agents & tools (48). At $0.755 per million tokens blended it is mid-priced. Output speed of 80 tokens per second puts it slower than most, with a first token in 2.5 s. It has been measured at 4 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 19 Oct 2026.

Blended price: $0.755/M; $0.380 in · $1.88 out
Output speed: 80 tok/s; OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
First answer: 2.48 s
Context: 1.0M
Full answer: –
Cost per run: –
Released: 13 Aug 2026

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

Overall index64
Reasoning67
Coding63
Agents & tools48
Maths64
Knowledge68
Human preference68
Multimodal68
Long context66
Composite73

Versions

Google has shipped 12 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
Gemini 3.8 Flashhigh	2 Sept 2026	64.5	#48
Gemini 3.7 Flashhighthis page	13 Aug 2026	63.5	#61
Gemini 3.6 Flashhigh	21 Jul 2026	62.1	#89
Gemini 3.5 Flashhigh	19 May 2026	63.2	#68
Gemini 3 Flashthinking	17 Dec 2025	59.2	#134
Gemini 2.5 Flash 09 2025thinking	25 Sept 2025	51.1	#348
Gemini 2.5 Flash	17 Apr 2025	52.7	#302
Gemini 2.0 Flash	5 Feb 2025	43.3	#560
Gemini 1.5 Flash 002	24 Sept 2024	40.2	#650
Gemini 1.5 Flash 001	23 May 2024	37.2	#726
Gemini 1.5 Flash	14 May 2024	–	–
Gemini Flash 2.0	–	–	–

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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
low	60.4	#112	80 tok/s	2.48 s	$0.0014	Agents & tools 58 · Coding 59 · Composite 71 · Knowledge 72 · Long context 64 · Multimodal 68 · Reasoning 55
medium	63.1	#71	80 tok/s	2.48 s	$0.0014	Agents & tools 64 · Coding 68 · Composite 74 · Knowledge 72 · Long context 67 · Multimodal 68 · Reasoning 60
highbest	63.5	#61	80 tok/s	2.48 s	$0.0007	Agents & tools 48 · Coding 63 · Composite 73 · Human preference 68 · Knowledge 68 · Long context 66 · Maths 64 · Multimodal 68 · Reasoning 67
not stated	53.2	#293	80 tok/s	2.48 s	$0.0014	Agents & tools 57 · Coding 57 · Composite 49 · Maths 58

How its index has moved

8 Sept 2026 to 20 Oct 2026

high
medium
low
not stated

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. 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	low	medium	high	not stated	Source
GPQA Diamond	–	–	94.8%#6	–	Epoch AI Benchmarking Hub
Arena Hard Prompts	–	–	1507#20	–	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	–	–	87.8%#23	LiveBench
LiveBench 1.0 Reasoningnot in index	–	–	87.8%#25	–	LiveBench
GPQA Diamond (AA)not in index	90.1%#64	92.1%#39	94.5%#6	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	35.1%#117	39.0%#94	47.9%#36	–	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	93.9%#5	–	Vals AI
ARC-AGI-1	85.2%#104	91.2%#61	95.5%#31	–	ARC Prize
ARC-AGI-2	52.9%#98	63.8%#72	84.6%#31	–	ARC Prize
CritPt	5.7%#145	9.4%#122	14.3%#97	–	Artificial Analysis
Chess Puzzlesnot in index	–	–	66.0%#6	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	–	–	37.0%#18	–	Epoch AI Benchmarking Hub
LMCAnot in index	47.1%#81	49.2%#72	50.4%#63	–	Epoch AI Benchmarking Hub
DTBenchnot in index	93.3%#64	94.7%#52	96.8%#21	–	Epoch AI Benchmarking Hub
ForecastBenchnot in index	–	–	66.2%#5	–	ForecastBench

Coding

Benchmark	low	medium	high	not stated	Source
SciCode	53.6%#72	59.8%#11	57.2%#32	–	SciCode
FrontierCodeeffort not stated	–	–	–	43.6%#18	Cognition
Arena Coding	–	–	1518#41	–	Arena (LMArena)
Arena WebDev	–	–	1591#29	–	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	–	–	80.5%#31	LiveBench
LiveBench 1.0 Codingnot in index	–	–	78.9%#29	–	LiveBench
SciCode (AA)not in index	55.7%#46	59.8%#12	57.2%#31	–	Artificial Analysis
LiveCodeBench	–	–	88.7%#5	–	Vals AI
IOI	–	–	67.8%#14	–	Vals AI
SWE-bench (Vals)not in index	–	–	80.8%#23	–	Vals AI
Code Migrationnot in index	–	–	34.8%#37	–	Vals AI
ProgramBenchnot in index	–	–	0.0%#32	–	Vals AI
Vibe Code Bench v1.1not in index	–	–	70.4%#40	–	Vals AI
DeepSWE v1.1not in index	53.8%#42	65.5%#25	65.3%#27	–	Epoch AI Benchmarking Hub
ALE-Benchnot in index	–	–	904.3#65	–	Epoch AI Benchmarking Hub
FrontierSWEnot in index	–	–	20.3%#15	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	–	–	13.6%#74	–	Artificial Analysis
Terminal-Bench 2.1 (AA)not in index	79.8%#46	78.3%#52	85.8%#24	–	Artificial Analysis

Agents & tools

Benchmark	low	medium	high	not stated	Source
Terminal-Bench	–	–	11.2%#82	–	Terminal-Bench
Remote Labor Indexeffort not stated	–	–	–	5.0%#8	Scale AI / CAIS
APEX-Agentseffort not stated	–	–	–	67.8%#5	Mercor
Arena Agent	–	–	-1.1#31	–	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	–	–	28.3%#33	LiveBench
LiveBench Browsernot in indexeffort not stated	–	–	–	56.0%#29	LiveBench
LiveBench Agentic Codingnot in index	–	–	58.3%#19	–	LiveBench
Terminal-Bench 2.1 (Vals)	–	–	77.5%#14	–	Vals AI
GDPval-AA v2.1not in index	41.2%#97	43.0%#85	44.6%#77	–	Artificial Analysis
τ³-Banking (AA)	29.5%#68	35.5%#46	32.8%#57	–	Artificial Analysis
Analyst Agent (AA)not in index	–	–	60.0%#1	–	Artificial Analysis
CyberBenchnot in index	–	–	43.8%#40	–	Vals AI
Finance Agent v2not in index	–	–	59.0%#7	–	Vals AI
Legal Research Benchnot in index	–	–	34.6%#42	–	Vals AI
SkillsBenchnot in index	–	–	65.9%#3	–	Vals AI
SREBenchnot in index	–	–	4.6%#17	–	Vals AI
Tax Agent Benchnot in index	–	–	22.8%#40	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	–	–	12.1%#35	–	Vals AI
Terminal-Bench Sciencenot in index	–	–	7.1%#20	–	Vals AI
AutomationBenchnot in index	–	–	62.0%#27	–	Artificial Analysis
GDP.pdf (AA)not in index	–	–	23.6%#42	–	Artificial Analysis
EnterpriseOps-Gymnot in index	–	50.4%#2	–	–	Artificial Analysis
FACTS Searchnot in indexeffort not stated	–	–	–	80.9%#12	Google

Maths

Benchmark	high	not stated	Source
FrontierMath Tiers 1–3	71.6%#26	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	36.6%#33	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	97.2%#36	–	Epoch AI Benchmarking Hub
ProofBencheffort not stated	–	58.0%#21	Vals AI
LiveBench Mathematicsnot in index	93.5%#22	–	LiveBench
Arena Maths	1506#9	–	Arena (LMArena)

Knowledge

Benchmark	low	medium	high	not stated	Source
SimpleQA Verified	–	–	69.2%#10	–	Epoch AI Benchmarking Hub
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	–	–	68.0%#36	LiveBench
LiveBench Researchnot in indexeffort not stated	–	–	–	76.5%#22	LiveBench
LiveBench 1.0 Data Analysisnot in index	–	–	68.0%#54	–	LiveBench
AA-Omniscience	22.1#57	23.7#54	26.5#48	–	Artificial Analysis
MMLU-Pro	–	–	90.1%#6	–	Vals AI
LegalBench	–	–	87.3%#6	–	Vals AI
TaxEval	–	–	74.7%#30	–	Vals AI
Excel Modeling Benchmarknot in index	–	–	71.3%#12	–	Vals AI
MedCodenot in index	–	–	53.4%#9	–	Vals AI
MedScribenot in index	–	–	83.9%#39	–	Vals AI
GDP.pdfnot in index	–	21.8%#29	23.8%#21	–	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	53.6%#46	54.0%#42	55.3%#37	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	32.3%#209	34.1%#194	35.5%#185	–	Artificial Analysis

Instruction following

Benchmark	high	Source
LiveBench Languagenot in index	85.5%#13	LiveBench
LiveBench Instruction Followingnot in index	79.9%#2	LiveBench
Arena Instruction Followingnot in index	1486#17	Arena (LMArena)

Human preference

Benchmark	high	Source
Arena Text	1488#17	Arena (LMArena)
Arena Creative Writingnot in index	1491#5	Arena (LMArena)
Arena Multi-turnnot in index	1490#22	Arena (LMArena)
Arena Longer Queriesnot in index	1499#17	Arena (LMArena)

Multimodal

Benchmark	low	medium	high	not stated	Source
Arena Vision	–	–	1314#8	–	Arena (LMArena)
LiveBench Designnot in indexeffort not stated	–	–	–	73.2%#32	LiveBench
MMMU-Pro	84.9%#14	84.7%#15	85.5%#10	–	Artificial Analysis
MMMU-Pro (Vals)not in index	–	–	89.0%#5	–	Vals AI
MortgageTaxnot in index	–	–	66.7%#32	–	Vals AI
SAGEnot in index	–	–	49.2%#23	–	Vals AI

Long context

Benchmark	low	medium	high	Source	Trend
AA-LCR	78.7%#109	83.0%#25	81.7%#51	Artificial Analysis
MLCRnot in index	–	–	15.0%#46	Artificial Analysis

Composite

Benchmark	low	medium	high	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	–	–	158.0#18	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	–	–	67.2%#29	LiveBench
LiveBench 1.0not in index	–	–	78.8%#16	–	LiveBench
AA Intelligence Index v4.3.2	37.0#84	39.6#66	39.1#71	–	Artificial Analysis

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

#5LiveCodeBench88.7% · of 137

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 Vertex	121 tok/s	2.18 s	$1.35	$6.75	1.0M	–
Google AI Studio	98 tok/s	2.36 s	$0.750	$3.75	1.0M	–
Google Vertex	88 tok/s	2.36 s	$0.750	$3.75	1.0M	–
Google Vertex	83 tok/s	1.56 s	$0.825	$4.13	1.0M	–
Google AI Studio	67 tok/s	1.69 s	$0.375	$1.88	1.0M	–
Google Vertex	67 tok/s	15 s	$0.375	$1.88	1.0M	–
Google AI Studio	56 tok/s	1.72 s	$1.35	$6.75	1.0M	–

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

input outputnow $0.750 in · $3.75 out

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 $0.075 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0007	$0.0006	6.2 s
Summarize a 30-page report	12,000 / 600	$0.0057	$0.0029	10.0 s
Code edit	6,000 / 1,500	$0.0051	$0.0037	21.2 s
Agentic coding session	60,000 / 4,000	$0.030	$0.017	52.5 s
Structured extraction	2,000 / 200	$0.0011	$0.0007	5.0 s