API host · catalogue checked 27 Oct 2026

Google: models, prices and uptime

Google serves 15 of the 78 models we benchmark through OpenRouter, 1 of them on no other host. Prices are OpenRouter's list prices per million tokens. We did not measure speed per host.

Models served15of 78 we test

Cheapest host for7of 14 shared models

Median premium+41%over the cheapest host

Uptime75.7%last 30 min, average

Most accurate models on Google

Verification score of each model this host serves

1Claude Opus 5.5Anthropic85.0%
2Gemini 4 ArgonGoogle84.8%
3Gemini 3.8 FlashGoogle82.5%
4Claude Fable 5.1Anthropic81.2%
5Claude Sonnet 5.5Anthropic78.8%
6Gemini 3.1 Pro PreviewGoogle77.5%
7Gemini 3.5 Flash LiteGoogle75.0%
8Claude Haiku 5.5Anthropic68.8%
9Llama 4 MaverickMeta59.0%
10gpt-oss-120bOpenAI58.8%
11gpt-oss-20bOpenAI51.2%
12Llama 3.3 70B InstructMeta42.5%
13Llama 4 ScoutMeta41.2%
14Gemma 4 26B A4BGoogle15.0%

Models on Google

Premium: this host's blended price (3 input : 1 output) against the cheapest host for the same model

Model	Score	Input / 1M	Output / 1M	vs cheapest host	Hosts	Context	Max output	Quantization	Uptime
Claude Opus 5.5	85.0%	$4	$20	Cheapest	5	1M	128k	—	100%
Gemini 4 Argon	84.8%	$2	$10	+100%	2	1M	944k	—	100%
Gemini 3.8 Flash	82.5%	$0.375	$1.875	Cheapest	2	1M	66k	—	26.7%
Claude Fable 5.1	81.2%	$10	$50	Cheapest	4	1M	128k	—	100%
Claude Sonnet 5.5	78.8%	$2	$10	Cheapest	5	1M	128k	—	100%
Gemini 3.1 Pro Preview	77.5%	$1	$6	Cheapest	2	1M	66k	—	100%
Gemini 3.5 Flash Lite	75.0%	$0.15	$1.25	Cheapest	2	1M	66k	—	100%
Claude Haiku 5.5	68.8%	$0.1	$0.5	Cheapest	5	1M	128k	—	100%
Llama 4 Maverick	59.0%	$0.35	$1.15	+81%	3	524k	8k	—	—
gpt-oss-120b	58.8%	$0.09	$0.36	+163%	23	131k	118k	—	16.7%
gpt-oss-20b	51.2%	$0.07	$0.25	+219%	10	131k	33k	—	0.0%
Llama 3.3 70B Instruct	42.5%	$0.72	$0.72	+365%	9	128k	8k	—	—
Llama 4 Scout	41.2%	$0.25	$0.7	+202%	2	1.3M	8k	—	—
Gemma 4 26B A4B	15.0%	$0.15	$0.6	+203%	15	262k	236k	—	89.7%
Qwen3 Next 80B A3B Thinking	—	$0.15	$1.2	Only host	1	262k	236k	—	—

All API hosts →

Google: frequently asked questions

Which AI models does Google serve?: Of the models we benchmark, Google serves 15 through OpenRouter: Claude Opus 5.5, Gemini 4 Argon, Gemini 3.8 Flash, Claude Fable 5.1, Claude Sonnet 5.5, Gemini 3.1 Pro Preview, Gemini 3.5 Flash Lite, Claude Haiku 5.5, Llama 4 Maverick, gpt-oss-120b, gpt-oss-20b, Llama 3.3 70B Instruct, Llama 4 Scout, Gemma 4 26B A4B, Qwen3 Next 80B A3B Thinking.
What is the most accurate model on Google?: Claude Opus 5.5 by Anthropic, with a verification score of 85.0% (68 of 80 questions correct) in our tests.
Is Google the cheapest host?: Google has the lowest list price for 7 of the 14 models it serves that other hosts also serve (Claude Opus 5.5, Gemini 3.8 Flash, Claude Fable 5.1, Claude Sonnet 5.5, Gemini 3.1 Pro Preview, Gemini 3.5 Flash Lite, Claude Haiku 5.5). Its median markup over the cheapest host is +41%.
How reliable is Google?: Averaged over the models it serves, Google reported 75.7% uptime in the 30 minutes before OpenRouter's catalogue was read for our run.

Model

Score

Input / 1M

Output / 1M

vs cheapest host

Hosts

Context

Max output

Quantization

Uptime

Claude Opus 5.5

85.0%

$20

Cheapest

128k

—

100%

Gemini 4 Argon

84.8%

$10

+100%

944k

—

100%

Gemini 3.8 Flash

82.5%

$0.375

$1.875

Cheapest

66k

—

26.7%

Claude Fable 5.1

81.2%

$10

$50

Cheapest

128k

—

100%

Claude Sonnet 5.5

78.8%

$10

Cheapest

128k

—

100%

Gemini 3.1 Pro Preview

77.5%

Cheapest

66k

—

100%

Gemini 3.5 Flash Lite

75.0%

$0.15

$1.25

Cheapest

66k

—

100%

Claude Haiku 5.5

68.8%

$0.1

$0.5

Cheapest

128k

—

100%

Llama 4 Maverick

59.0%

$0.35

$1.15

+81%

524k

—

gpt-oss-120b

58.8%

$0.09

$0.36

+163%

131k

118k

—

16.7%

gpt-oss-20b

51.2%

$0.07

$0.25

+219%

131k

33k

—

0.0%

Llama 3.3 70B Instruct

42.5%

$0.72

+365%

128k

—

Llama 4 Scout

41.2%

$0.25

$0.7

+202%

1.3M

—

Gemma 4 26B A4B

15.0%

$0.15

$0.6

+203%

262k

236k

—

89.7%

Qwen3 Next 80B A3B Thinking

—

$0.15

$1.2

Only host

262k

236k

—

Google: frequently asked questions

Which AI models does Google serve?

Of the models we benchmark, Google serves 15 through OpenRouter: Claude Opus 5.5, Gemini 4 Argon, Gemini 3.8 Flash, Claude Fable 5.1, Claude Sonnet 5.5, Gemini 3.1 Pro Preview, Gemini 3.5 Flash Lite, Claude Haiku 5.5, Llama 4 Maverick, gpt-oss-120b, gpt-oss-20b, Llama 3.3 70B Instruct, Llama 4 Scout, Gemma 4 26B A4B, Qwen3 Next 80B A3B Thinking.

What is the most accurate model on Google?

Claude Opus 5.5 by Anthropic, with a verification score of 85.0% (68 of 80 questions correct) in our tests.

Is Google the cheapest host?

Google has the lowest list price for 7 of the 14 models it serves that other hosts also serve (Claude Opus 5.5, Gemini 3.8 Flash, Claude Fable 5.1, Claude Sonnet 5.5, Gemini 3.1 Pro Preview, Gemini 3.5 Flash Lite, Claude Haiku 5.5). Its median markup over the cheapest host is +41%.

How reliable is Google?

Averaged over the models it serves, Google reported 75.7% uptime in the 30 minutes before OpenRouter's catalogue was read for our run.