A · closed · Gemini

Gemini 1.5 Flash-8B

Google

B · closed · Gemini

Gemini 3.6 Flash

Google

Replace A

Replace B

0:6

overall capability wins (1 ties across 7 scored). Gemini 3.6 Flash leads overall.

Reasoning0:3
Coding0:1
Arena0:1
Pricing3:0
Speed1:0
Specs0:1

Verdict

Gemini 3.6 Flash leads coding (widest gap: +30.5 on SciCode); Gemini 3.6 Flash ranks higher on LMArena (+221 Elo); Gemini 3.6 Flash allows ~8.0x more max output (66K vs 8K).

Gemini 3.6 Flash leads coding (widest gap: +30.5 on SciCode)
Gemini 3.6 Flash ranks higher on LMArena (+221 Elo)
Gemini 3.6 Flash allows ~8.0x more max output (66K vs 8K)
Gemini 1.5 Flash-8B is ~45.7x cheaper on a blended token basis than Gemini 3.6 Flash

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

Gemini 1.5 Flash-8B: $0.07
Gemini 3.6 Flash: $3.38

Gemini 1.5 Flash-8B is about 45.0x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

Google

Family

Gemini

License

Proprietary

Open weights

Release

Oct 3, 2024

Jul 21, 2026

Knowledge cutoff

API / provider

Google

Modalities

text, image, audio, video → text

Specs

Context window

tie

Max output

66K

B +57K

Parameters

—

Pricing

Input $/1M

$0.04

$1.50

A +$1.46

Output $/1M

$0.15

$7.50

A +$7.35

Blended $/1M (3∶1)

$0.07

A +$2.93

Speed

tok/s

180

tie

TTFT (s)

0.2

0.22

A +0.02

Reasoning

MMLU-Pro

56.9

—

GPQA Diamond

35.9

92.8

B +56.9 pts

Humanity's Last Exam

4.7

40.8

B +36.1 pts

AIME 2025

—

MATH-500

68.9

—

Humanity's Last Exam (with tools)

—

AA-Omniscience Accuracy

—

50.0

—

AA-LCR v1.1

—

80.0

—

CritPt

—

10.6

—

MMMU-Pro

36.5

83.2

B +46.7 pts

IFBench

—

Chartography

—

Chartography (With Tools)

—

Coding

SWE-bench Verified

—

SWE-bench Pro

—

SWE-bench Multilingual

—

LiveCodeBench

21.7

—

Terminal-Bench 2.1

—

77.5

—

Aider Polyglot

—

Terminal-Bench 3

—

BigCodeBench

—

SciCode

22.9

53.4

B +30.5 pts

CursorBench

—

53.5

—

SWE-Rebench

—

NL2Repo-Bench

—

DeepSWE

—

47.0

—

WebDev Arena

—

1,537

—

Terminal-Bench 4.0

—

CursorBench 4.0

—

FrontierCode v1.1 (Main)

—

Arena

LMArena Elo

1,259

1,480

B +221 Elo

Metric	Gemini 1.5 Flash-8B	Gemini 3.6 Flash	Delta
Identity
Organization	Google	Google	—
Family	Gemini	Gemini	—
License	Proprietary	Proprietary	—
Open weights	No	No	—
Release	Oct 3, 2024	Jul 21, 2026	—
Knowledge cutoff	-	-	—
API / provider	Google	Google	—
Modalities	text, image, audio, video → text	text, image, audio, video → text	—
Specs
Context window	1M	1M	tie
Max output	8K	66K	B +57K
Parameters	—	—	—
Pricing
Input $/1M	$0.04	$1.50	A +$1.46
Output $/1M	$0.15	$7.50	A +$7.35
Blended $/1M (3∶1)	$0.07	$3	A +$2.93
Speed
tok/s	180	180	tie
TTFT (s)	0.2	0.22	A +0.02
Reasoning
MMLU-Pro	56.9	—	—
GPQA Diamond	35.9	92.8	B +56.9 pts
Humanity's Last Exam	4.7	40.8	B +36.1 pts
AIME 2025	—	—	—
MATH-500	68.9	—	—
Humanity's Last Exam (with tools)	—	—	—
AA-Omniscience Accuracy	—	50.0	—
AA-LCR v1.1	—	80.0	—
CritPt	—	10.6	—
MMMU-Pro	36.5	83.2	B +46.7 pts
IFBench	—	—	—
Chartography	—	—	—
Chartography (With Tools)	—	—	—
Coding
SWE-bench Verified	—	—	—
SWE-bench Pro	—	—	—
SWE-bench Multilingual	—	—	—
LiveCodeBench	21.7	—	—
Terminal-Bench 2.1	—	77.5	—
Aider Polyglot	—	—	—
Terminal-Bench 3	—	—	—
BigCodeBench	—	—	—
SciCode	22.9	53.4	B +30.5 pts
CursorBench	—	53.5	—
SWE-Rebench	—	—	—
NL2Repo-Bench	—	—	—
DeepSWE	—	47.0	—
WebDev Arena	—	1,537	—
Terminal-Bench 4.0	—	—	—
CursorBench 4.0	—	—	—
FrontierCode v1.1 (Main)	—	—	—
Arena
LMArena Elo	1,259	1,480	B +221 Elo