Models
Every model the API serves: price per 1M tokens, context window and capabilities.
Updated Nov 19, 2026, 4:17 PM
- GLM 5.3glm1M context131.1K max outputInput€0.093Cached input€0.092Output€1.90
- DeepSeek V4.1 Flashdeepseek1M context393.2K max outputInput€0.11Cached input€0.014Output€0.43
- DeepSeek DGXdeepseek1M context393.2K max outputInput€0.067Cached input€0.0067Output€0.27
- GLM 5.3 Flashglm1M context131.1K max outputInput€0.18Cached input€0.04Output€0.59
Two serving tiers, labelled
Partner routes are admitted only under a written zero-retention agreement. Rented GPU routes serve the same models on machines the platform rents, and are labelled transient: the platform stores nothing there either, but the machine operator holds root and can see a request while it runs. The price of a request is the same on either tier. A key set to zero-retention routes only, or a model id with the :zdr suffix, is served only by the partner tier.
Zero retention, in fullAbliterated variants
A model whose id ends in -abliterated has the refusal behaviour the provider trained into it removed. It is served under the same retention rules and the same acceptable use policy as every other model.