Supported models
Modal Endpoints serve the open-weight models listed on this page. Shared Endpoints serve models from a Modal-managed compute pool, using token-based billing. Dedicated Endpoints run the models listed in Dedicated Endpoint models, or compatible custom weights, on isolated, autoscaling containers.
Learn more about model capabilities in the Model Library.
Shared Endpoint models
Shared Endpoints are billed per token. Rates are in USD per million tokens and also appear when you create an Endpoint.
| Model | Input Modalities | Input ($/Mtok) | Output ($/Mtok) |
|---|---|---|---|
| GLM 5.3 Flash NVFP4 | Text, Image | 0.15 / 0.03 cached | 0.50 |
| DeepSeek V4.1 Flash 1017 | Text, Image | 0.25 / 0.005 cached | 1.00 |
| DeepSeek V4.1 Flash | Text, Image | 0.30 / 0.03 cached | 1.20 |
| GLM 5.3 Flash | Text, Image | 0.45 / 0.09 cached | 1.50 |
| Muse Spark Open | Text, Image, Audio | 1.25 / 0.15 cached | 4.25 |
| GLM 5.3 | Text | 1.40 / 0.26 cached | 4.40 |
| Inkling NVFP4 | Text, Image, Audio | 1.20 / 0.27 cached | 5.00 |
| Qwen3.8-Max | Text | 2.00 / 0.25 cached | 6.00 |
| Qwen3.8-Max VL | Text, Image | 2.00 / 0.25 cached | 6.00 |
| Kimi K3 | Text, Image | 3.00 / 0.30 cached | 15.00 |
Dedicated Endpoint models
Dedicated Endpoints bill for compute required to serve the model; see Dedicated Endpoint pricing.
For entries shown as model IDs, pass the ID to modal endpoint create --model.