IBM watsonx
watsonx.ai is IBM's platform for building with models. It sells IBM's Granite models and others per token, and runs models deployed on demand on dedicated GPUs by the hour.
- Website: https://www.ibm.com/products/watsonx-ai
- Headquarters: Armonk, United States
- Kind: Enterprise AI platform
IBM watsonx sells 3 of the compared models per token and prices 4 kinds of GPU for hosting your own.
Models sold per token
| Model | Input, per 1M tokens | Output, per 1M tokens |
|---|---|---|
| Llama 3.3 70B | $0.7526 | $0.7526 |
| Llama 4 Maverick 17B | $0.371 | $1.48 |
| OpenAI gpt-oss 120B | $0.159 | $0.636 |
GPUs for hosting your own
| GPU | From, per GPU-hour | Cheapest instance | GPUs in it |
|---|---|---|---|
| L40S | $4.43 | deploy-on-demand.1xL40S | 1 × 48 GB |
| A100 | $5.80 | deploy-on-demand.1xA100 | 1 × 80 GB |
| H100 | $14.50 | deploy-on-demand.1xH100 | 1 × 80 GB |
| H200 | $16.00 | deploy-on-demand.1xH200 | 1 × 141 GB |
Where it runs
One global price; IBM watsonx chooses where a request is processed.
No place stated for 3 models: one global price, and IBM watsonx does not say where a request is processed.
Where the prices come from
- Per-token prices: dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models.html
- GPU servers: dataplatform.cloud.ibm.com/docs/content/wsj/getting-started/wxai-runtime-plans-genai.html
Sources
List prices collected 4 November 2026, refreshed every three hours. No negotiated rates or taxes.