| Model | Type | Price, $/1M | Context | Released |
|---|---|---|---|---|
| Granite 4.1 8B | text | $0.06 | 131K | Apr 2026 |
| Granite 4.0 H Micro | text | $0.04 | 131K | Oct 2025 |
| Granite 13B Chat V2 | text | $0.6 | 8K | — |
| Granite 13B Instruct V2 | text | $0.6 | 8K | — |
| Granite 3.3 8B Instruct | text | $0.09 | 8K | — |
| Granite 3 8B Instruct | text | $0.2 | 8K | — |
| Granite 4 H Small | text | $0.11 | 20K | — |
| Granite Guardian 3.2 2B | moderation | $0.1 | 8K | — |
| Granite Guardian 3.3 8B | moderation | $0.2 | 8K | — |
| Granite Ttm 1024.96 R2 | text | $0.38 | 1K | — |
| Granite Ttm 1536.96 R2 | text | $0.38 | 1K | — |
| Granite Ttm 512.96 R2 | text | $0.38 | 1K | — |
| Granite Vision 3.2 2B | text | $0.1 | 8K | — |
A blended price, on a scheme of three input tokens per one output token. It is taken from the developer itself, if they sell this model; if not, from the cheapest offer among providers, as a pair from one price list. Base per-token tier only. A dash means the model has no per-token price in open sources, not that it is free.