LLM Opex

Vendors

Alibaba Cloud

Alibaba Cloud is Alibaba Group's cloud. It sells its Qwen models and others per token through Model Studio and rents GPU instances by the hour on ECS, outside mainland China here.

Alibaba Cloud sells 8 of the compared models per token and prices 5 kinds of GPU for hosting your own.

Models sold per token

The cheapest price anywhere, at the standard tier.
ModelInput, per 1M tokensOutput, per 1M tokens
DeepSeek V4 Pro$1.65$3.30
DeepSeek V4.1 Flash$0.283$1.13
GLM 5.1$0.825$3.30
GLM 5.2$1.10$3.85
GLM 5.3$1.13$3.96
Kimi K2.5$0.574$3.01
Kimi K2.7 Code$0.894$3.71
Kimi K3$2.83$14.13

GPUs for hosting your own

On demand.
GPUFrom, per GPU-hourCheapest instanceGPUs in it
T4$1.12ecs.gn6i-c4g1.xlarge1 × 16 GB
A10$1.27ecs.gn7i-4x.8xlarge4 × 24 GB
L20$2.25ecs.gn8is-8x.32xlarge8 × 48 GB
L20N$6.71ecs.gn9gc.4xlarge1 × 72 GB
GPU H$7.19ecs.gn8v-2x.8xlarge2 × 96 GB

Where it runs

10 regions in 9 countries.

CountryRegionsPer-token modelsGPU kinds
Germanyeu-central-183
United Statesus-east-1, us-west-184
Japanap-northeast-173
Singaporeap-southeast-173
Hong Kongcn-hongkong43
Indonesiaap-southeast-5—3
Malaysiaap-southeast-3—2
Saudi Arabiame-central-1—1
Thailandap-southeast-7—2

Where the prices come from

Sources

List prices collected 4 November 2026, refreshed every three hours. No negotiated rates or taxes.