GLM 5.3 Flash

Z.ai2026-08-26● 开源 · MIT🤗 zai-org/GLM-5.3-Flash648.9万2,823

GLM-5.3-Flash 是智谱（Z.ai）推出的原生多模态模型，适用于高效编码与长周期 Agent 任务，其混合稀疏与线性注意力架构在保持长上下文准确性方面表现出色。

多服务商价格对比数据来自 OpenRouter

32 个服务商1 USD ≈ 6.7058 CNY · 更新于 2026-10-13

服务商	量化精度	上下文	输入/M	输出/M	缓存读取/M	近1天可用率
O OpenInference	fp4	1.05M	¥0.18	¥6.04	¥0.07	98.8%
Morph	fp8	1.05M	¥0.19	¥6.71	¥0.10	94.1%
Relace	—	1.05M	¥0.27	¥3.35	¥0.08	100.0%
StreamLake限时 -71%	fp8	1.02M	¥0.29	¥0.97	¥0.06	99.9%
S Sail Research	fp4	1.05M	¥0.30	¥4.02	¥0.19	99.9%
S Sail Research	fp4	1.05M	¥0.30	¥4.02	¥0.19	100.0%
R Reka	—	262K	¥0.40	¥10.73	¥0.27	100.0%
I InferenceNet	fp4	1.05M	¥0.47	¥0.34	¥0.20	100.0%
DeepInfra限时 -50%	fp4	1.05M	¥0.50	¥1.68	¥0.10	-
W Wafer	—	1.05M	¥0.50	¥3.35	¥0.08	100.0%
I Inceptron	fp8	1.05M	¥0.53	¥3.69	¥0.66	99.3%
Novita限时 -44%	fp8	1.05M	¥0.56	¥1.88	¥0.11	100.0%
G GMICloud限时 -40%	fp8	1.05M	¥0.60	¥2.01	¥0.12	98.6%
D Decart限时 -38%	fp4	1.05M	¥0.62	¥2.08	¥0.12	100.0%
N Near AI限时 -30%	fp8	1.05M	¥0.70	¥2.35	¥0.16	100.0%
P Phala限时 -25%	nvfp4	1.05M	¥0.75	¥2.51	¥0.15	99.9%
DigitalOcean	—	1.05M	¥0.88	¥2.93	¥0.20	99.0%
Modal	nvfp4	1.05M	¥1.01	¥3.35	¥0.20	100.0%
BaseTen	fp8	1.05M	¥1.01	¥3.35	¥0.20	100.0%
Crusoe	fp4	1.05M	¥1.01	¥3.35	¥0.20	100.0%
C CoreWeave	nvfp4	1.05M	¥1.01	¥3.35	¥0.34	99.3%
AtlasCloud	fp8	1.05M	¥1.01	¥3.35	¥0.20	94.2%
Fireworks	—	1.05M	¥1.01	¥3.35	¥0.20	100.0%
Friendli	—	1.05M	¥1.01	¥3.35	¥0.20	100.0%
S SiliconFlow	fp8	1.05M	¥1.01	¥3.35	¥0.20	90.6%
Together	—	1.05M	¥1.01	¥3.35	¥0.20	100.0%
Parasail	fp4	1.05M	¥1.01	¥3.35	¥0.20	100.0%
BaseTen	fp8	1.05M	¥1.01	¥3.35	¥0.20	100.0%
Venice	—	1.05M	¥1.01	¥3.35	¥0.20	100.0%
Z.AI	fp8	1.05M	¥1.01	¥3.35	¥0.20	100.0%
Parasail	fp4	1.05M	¥1.26	¥4.19	¥0.25	99.9%
Fireworks (US)	—	1.05M	¥1.51	¥5.03	¥0.30	100.0%

O

OpenInferencefp4

¥0.18

输入/M

¥6.04

输出/M

¥0.07

缓存/M

98.8%

可用率

Morphfp8

¥0.19

输入/M

¥6.71

输出/M

¥0.10

缓存/M

94.1%

可用率

Relace

¥0.27

输入/M

¥3.35

输出/M

¥0.08

缓存/M

100.0%

可用率

StreamLakefp8限时 -71%

¥0.29

输入/M

¥0.97

输出/M

¥0.06

缓存/M

99.9%

可用率

S

Sail Researchfp4

¥0.30

输入/M

¥4.02

输出/M

¥0.19

缓存/M

99.9%

可用率

S

Sail Researchfp4

¥0.30

输入/M

¥4.02

输出/M

¥0.19

缓存/M

100.0%

可用率

R

Reka

¥0.40

输入/M

¥10.73

输出/M

¥0.27

缓存/M

100.0%

可用率

I

InferenceNetfp4

¥0.47

输入/M

¥0.34

输出/M

¥0.20

缓存/M

100.0%

可用率

DeepInfrafp4限时 -50%

¥0.50

输入/M

¥1.68

输出/M

¥0.10

缓存/M

-

可用率

W

Wafer

¥0.50

输入/M

¥3.35

输出/M

¥0.08

缓存/M

100.0%

可用率

I

Inceptronfp8

¥0.53

输入/M

¥3.69

输出/M

¥0.66

缓存/M

99.3%

可用率

Novitafp8限时 -44%

¥0.56

输入/M

¥1.88

输出/M

¥0.11

缓存/M

100.0%

可用率

G

GMICloudfp8限时 -40%

¥0.60

输入/M

¥2.01

输出/M

¥0.12

缓存/M

98.6%

可用率

D

Decartfp4限时 -38%

¥0.62

输入/M

¥2.08

输出/M

¥0.12

缓存/M

100.0%

可用率

N

Near AIfp8限时 -30%

¥0.70

输入/M

¥2.35

输出/M

¥0.16

缓存/M

100.0%

可用率

P

Phalanvfp4限时 -25%

¥0.75

输入/M

¥2.51

输出/M

¥0.15

缓存/M

99.9%

可用率

DigitalOcean

¥0.88

输入/M

¥2.93

输出/M

¥0.20

缓存/M

99.0%

可用率

Modalnvfp4

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

BaseTenfp8

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

Crusoefp4

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

C

CoreWeavenvfp4

¥1.01

输入/M

¥3.35

输出/M

¥0.34

缓存/M

99.3%

可用率

AtlasCloudfp8

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

94.2%

可用率

Fireworks

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

Friendli

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

S

SiliconFlowfp8

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

90.6%

可用率

Together

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

Parasailfp4

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

BaseTenfp8

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

Venice

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

Z.AIfp8

¥1.01

输入/M

¥3.35

输出/M

¥0.20

缓存/M

100.0%

可用率

Parasailfp4

¥1.26

输入/M

¥4.19

输出/M

¥0.25

缓存/M

99.9%

可用率

Fireworks (US)

¥1.51

输入/M

¥5.03

输出/M

¥0.30

缓存/M

100.0%

可用率