diff --git a/apps/docs/models.mdx b/apps/docs/models.mdx
index bed4d385bc..621488b30e 100644
--- a/apps/docs/models.mdx
+++ b/apps/docs/models.mdx
@@ -157,23 +157,25 @@ when you want to use it.
 **GPT 5.6 Luna** is the default coding model for Roomote's default OpenRouter
 setup and for new Vercel AI Gateway, Requesty, OpenAI, Azure OpenAI, Azure AI
 Foundry, OpenCode Zen, OpenCode Go, Amazon Bedrock, GitHub Copilot, and ChatGPT
-subscription connections. The recommended OpenAI, ChatGPT subscription, and
-GitHub Copilot mapping presets also select GPT 5.6 Luna with medium reasoning.
-Existing saved defaults and role mappings do not change automatically.
+subscription connections. Where a provider offers a **Default** mapping preset,
+it uses that provider's GPT 5.6 Luna route for every role except orchestration,
+which follows the coding model, and uses medium reasoning for coding. Existing
+saved defaults and role mappings do not change automatically.
 
-Providers also carry recommended defaults for the model roles below — for
+Providers can also offer differentiated presets for the model roles below — for
 example a strong model for planning and code review and a fast, low-cost model
-for helper and explore work. Connecting a provider in the setup wizard applies
-its recommended defaults automatically, so a fresh deployment starts with a
-sensible split instead of one model for everything.
+for helper and explore work. The **Recommended** OpenAI and ChatGPT subscription
+presets use GPT-6 Sol for coding, while the GitHub Copilot preset uses GPT-6
+Luna. Connecting one of these providers in the setup wizard applies its
+**Default** preset; choose a differentiated preset later when you want to split
+roles across models.
 
 From **Settings > Models**, use **Use a mapping preset** on the **Model
-mapping** card to preview and apply a connected provider's recommended model
-mapping. Confirm the preset to set the role selections and enable any missing
-recommended models. It is an apply-once action, not a lock: you can adjust
-every role afterward. Roles a provider has no specific recommendation for are
-set to **Same as coding model**, and roles managed by environment variables are
-left untouched.
+mapping** card to preview and apply any preset offered by a connected provider.
+Confirm the preset to set the role selections and enable any missing models. It
+is an apply-once action, not a lock: you can adjust every role afterward. Roles
+a provider has no specific recommendation for are set to **Same as coding
+model**, and roles managed by environment variables are left untouched.
 
 You can also choose **Add your own** to save a private preset with a model and
 reasoning level for each role when that model supports reasoning. Your custom