# Helper jobs now run on your own key 2026-10-27 · Product > With your own provider key, summaries, /goal checks, Deep Research helpers and similar jobs now use a small model of that provider instead of a platform model. With your own provider key, the small jobs around a chat now run on a small model of the same provider. Before, several of them quietly switched to a platform model or failed. *Helper jobs of a chat on your own key use the default model of your provider, or the one you choose.* Text version Summaries of long chats, /goal checks, Deep Research helpers, memory suggestions, the prompt improver, the Ultracode plan and its checks, and second opinions from another model run on the default model of your provider. Your provider bills them. Platform chats keep their helper. The default is a small, fast model of the provider you connected, for example Claude Haiku 5.5 for Anthropic. These calls run on your key, so they need no ZeroLog credit. A chat on a platform model is unchanged. You can pick another model for each connected provider under Settings, Advanced, "Helper model with your own key". The model that describes pictures for text-only models can be set there too. Fixed on the way: an Ultracode checker that does not exist at your provider, or one that only turns text into numbers, is refused at the start instead of failing in the middle of a run. The model-to-model delegation and the Ultracode swarm now see your own models too. We tested this with a replica of the providers and the database. We have not yet run every helper against a real key of each of the 25 providers. Anthropic and the OpenAI style endpoints share the code that was tested.