API host · catalogue checked 27 Oct 2026

CoreWeave: models, prices and uptime

CoreWeave serves 15 of the 78 models we benchmark through OpenRouter. Prices are OpenRouter's list prices per million tokens. We did not measure speed per host.

Models served15of 78 we test

Cheapest host for2of 15 shared models

Median premium+73%over the cheapest host

Uptime99.6%last 30 min, average

Most accurate models on CoreWeave

Verification score of each model this host serves

1DeepSeek V4 Pro 0813DeepSeek71.2%
2GLM 5.3 FlashZ.ai67.5%
3GLM 5.3Z.ai67.5%
4Kimi K2.7 CodeMoonshotAI65.0%
5Qwen3.8 27BQwen63.8%
6gpt-oss-120bOpenAI58.8%
7MiniMax M3MiniMax57.5%
8gpt-oss-20bOpenAI51.2%
9Nemotron 3.5 LightningNVIDIA43.8%
10Llama 3.3 70B InstructMeta42.5%
11Nemotron 3.5 Super VLNVIDIA40.0%
12Qwen3.6 35B A3BQwen37.5%
13Granite 4.2 8BIBM36.2%
14Gemma 4 31BGoogle30.0%
15Gemma 4 26B A4BGoogle15.0%

Models on CoreWeave

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
DeepSeek V4 Pro 0813	71.2%	$1.31	$3.96	+840%	23	1M	944k	fp8	100%
GLM 5.3 Flash	67.5%	$0.12	$0.38	+294%	34	1M	944k	nvfp4	100%
GLM 5.3	67.5%	$1.2	$4	+372%	37	1M	944k	fp8	94.7%
Kimi K2.7 Code	65.0%	$0.6	$3	+0%	11	262k	236k	int4	100%
Qwen3.8 27B	63.8%	$0.4	$3	+518%	21	262k	236k	fp8	100%
gpt-oss-120b	58.8%	$0.03	$0.17	+8%	23	131k	118k	fp4	100%
MiniMax M3	57.5%	$0.23	$0.96	+6%	12	262k	236k	fp4	100%
gpt-oss-20b	51.2%	$0.03	$0.13	+53%	10	131k	118k	fp4	100%
Nemotron 3.5 Lightning	43.8%	$0.07	$0.2	+110%	4	262k	236k	bf16	100%
Llama 3.3 70B Instruct	42.5%	$0.71	$0.71	+358%	9	128k	115k	fp16	100%
Nemotron 3.5 Super VL	40.0%	$0.15	$0.5	+79%	9	1M	944k	nvfp4	100%
Qwen3.6 35B A3B	37.5%	$0.1	$0.95	+47%	10	262k	236k	fp8	100%
Granite 4.2 8B	36.2%	$0.06	$0.15	Cheapest	2	131k	118k	bf16	100%
Gemma 4 31B	30.0%	$0.1	$0.34	Cheapest	13	262k	236k	fp4	100%
Gemma 4 26B A4B	15.0%	$0.1	$0.3	+73%	15	262k	236k	bf16	100%

All API hosts →

CoreWeave: frequently asked questions

Which AI models does CoreWeave serve?: Of the models we benchmark, CoreWeave serves 15 through OpenRouter: DeepSeek V4 Pro 0813, GLM 5.3 Flash, GLM 5.3, Kimi K2.7 Code, Qwen3.8 27B, gpt-oss-120b, MiniMax M3, gpt-oss-20b, Nemotron 3.5 Lightning, Llama 3.3 70B Instruct, Nemotron 3.5 Super VL, Qwen3.6 35B A3B, Granite 4.2 8B, Gemma 4 31B, Gemma 4 26B A4B.
What is the most accurate model on CoreWeave?: DeepSeek V4 Pro 0813 by DeepSeek, with a verification score of 71.2% (57 of 80 questions correct) in our tests.
Is CoreWeave the cheapest host?: CoreWeave has the lowest list price for 2 of the 15 models it serves that other hosts also serve (Granite 4.2 8B, Gemma 4 31B). Its median markup over the cheapest host is +73%.
How reliable is CoreWeave?: Averaged over the models it serves, CoreWeave reported 99.6% 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

DeepSeek V4 Pro 0813

71.2%

$1.31

$3.96

+840%

944k

fp8

100%

GLM 5.3 Flash

67.5%

$0.12

$0.38

+294%

944k

nvfp4

100%

GLM 5.3

67.5%

$1.2

+372%

944k

fp8

94.7%

Kimi K2.7 Code

65.0%

$0.6

+0%

262k

236k

int4

100%

Qwen3.8 27B

63.8%

$0.4

+518%

262k

236k

fp8

100%

gpt-oss-120b

58.8%

$0.03

$0.17

+8%

131k

118k

fp4

100%

MiniMax M3

57.5%

$0.23

$0.96

+6%

262k

236k

fp4

100%

gpt-oss-20b

51.2%

$0.03

$0.13

+53%

131k

118k

fp4

100%

Nemotron 3.5 Lightning

43.8%

$0.07

$0.2

+110%

262k

236k

bf16

100%

Llama 3.3 70B Instruct

42.5%

$0.71

+358%

128k

115k

fp16

100%

Nemotron 3.5 Super VL

40.0%

$0.15

$0.5

+79%

944k

nvfp4

100%

Qwen3.6 35B A3B

37.5%

$0.1

$0.95

+47%

262k

236k

fp8

100%

Granite 4.2 8B

36.2%

$0.06

$0.15

Cheapest

131k

118k

bf16

100%

Gemma 4 31B

30.0%

$0.1

$0.34

Cheapest

262k

236k

fp4

100%

Gemma 4 26B A4B

15.0%

$0.1

$0.3

+73%

262k

236k

bf16

100%

CoreWeave: frequently asked questions

Which AI models does CoreWeave serve?

Of the models we benchmark, CoreWeave serves 15 through OpenRouter: DeepSeek V4 Pro 0813, GLM 5.3 Flash, GLM 5.3, Kimi K2.7 Code, Qwen3.8 27B, gpt-oss-120b, MiniMax M3, gpt-oss-20b, Nemotron 3.5 Lightning, Llama 3.3 70B Instruct, Nemotron 3.5 Super VL, Qwen3.6 35B A3B, Granite 4.2 8B, Gemma 4 31B, Gemma 4 26B A4B.

What is the most accurate model on CoreWeave?

DeepSeek V4 Pro 0813 by DeepSeek, with a verification score of 71.2% (57 of 80 questions correct) in our tests.

Is CoreWeave the cheapest host?

CoreWeave has the lowest list price for 2 of the 15 models it serves that other hosts also serve (Granite 4.2 8B, Gemma 4 31B). Its median markup over the cheapest host is +73%.

How reliable is CoreWeave?

Averaged over the models it serves, CoreWeave reported 99.6% uptime in the 30 minutes before OpenRouter's catalogue was read for our run.