diff --git a/packages/sdk/AGENTS.md b/packages/sdk/AGENTS.md
index 32cc5f9..acf25dd 100644
--- a/packages/sdk/AGENTS.md
+++ b/packages/sdk/AGENTS.md
@@ -79,7 +79,7 @@
 - `lucy-vton-3.5` - Real-time virtual try-on 3.5
 - `lucy-vton-3.6` - Real-time virtual try-on 3.6
 - `lucy-restyle-2` - Real-time video restyling
-- `lucy-latest`, `lucy-vton-latest`, `lucy-restyle-latest` - Server-resolved aliases for the latest stable version (`lucy-vton-latest` → `lucy-vton-3.6`)
+- `lucy-latest`, `lucy-vton-latest`, `lucy-restyle-latest` - Server-resolved aliases for the latest stable version (`lucy-latest` → `lucy-2.6`, `lucy-vton-latest` → `lucy-vton-3.6`)
 
 ### Video Models (Queue API)
 - `lucy-2.1` - long-form video editing (720p)
@@ -88,7 +88,7 @@
 - `lucy-vton-3.5` - virtual try-on 3.5 video editing
 - `lucy-vton-3.6` - virtual try-on 3.6 video editing
 - `lucy-restyle-2` - video restyling
-- `lucy-latest`, `lucy-vton-latest`, `lucy-restyle-latest` - Server-resolved aliases (`lucy-vton-latest` → `lucy-vton-3.6`)
+- `lucy-latest`, `lucy-vton-latest`, `lucy-restyle-latest` - Server-resolved aliases (`lucy-latest` → `lucy-2.6`, `lucy-vton-latest` → `lucy-vton-3.6`)
 
 ### Image Models (Process API)
 - `lucy-image-2` - image-to-image editing
diff --git a/packages/sdk/README.md b/packages/sdk/README.md
index 2d5bfde..c48977f 100644
--- a/packages/sdk/README.md
+++ b/packages/sdk/README.md
@@ -93,7 +93,7 @@ iOS does not show the microphone permission prompt that playing out any remote a
 
 #### Fast mode
 
-Fast mode (`speed: "fast"`) serves the session from a higher-compute tier for lower latency and higher throughput; output quality is unchanged. It is currently available for `lucy-2.6`, `lucy-2.5` / `lucy-latest` and `lucy-vton-3.6` / `lucy-vton-3.5` / `lucy-vton-latest`, in the US region only, and is billed at 2x the standard realtime rate for those models. Other models ignore the option. Omit it (the default) for standard mode.
+Fast mode (`speed: "fast"`) serves the session from a higher-compute tier for lower latency and higher throughput; output quality is unchanged. It is currently available for `lucy-2.6` / `lucy-latest`, `lucy-2.5` and `lucy-vton-3.6` / `lucy-vton-3.5` / `lucy-vton-latest`, in the US region only, and is billed at 2x the standard realtime rate for those models. Other models ignore the option. Omit it (the default) for standard mode.
 
 ```ts
 const realtimeClient = await client.realtime.connect(stream, {
diff --git a/packages/sdk/src/realtime/client.ts b/packages/sdk/src/realtime/client.ts
index f77d725..2cc9404 100644
--- a/packages/sdk/src/realtime/client.ts
+++ b/packages/sdk/src/realtime/client.ts
@@ -103,8 +103,8 @@ const realTimeClientConnectOptionsSchema = z.object({
   resolution: z.enum(["720p", "1080p"]).optional(),
   /**
    * Realtime speed tier. Fast mode (`"fast"`) serves the session from a higher-compute tier for lower
-   * latency and higher throughput; output quality is unchanged. It is currently available for `lucy-2.6`,
-   * `lucy-2.5` / `lucy-latest` and `lucy-vton-3.6` / `lucy-vton-3.5` / `lucy-vton-latest`, in the US region
+   * latency and higher throughput; output quality is unchanged. It is currently available for `lucy-2.6` /
+   * `lucy-latest`, `lucy-2.5` and `lucy-vton-3.6` / `lucy-vton-3.5` / `lucy-vton-latest`, in the US region
    * only, and is billed at 2x the standard realtime rate for those models. Other models ignore the option (the
    * SDK logs a warning when the model does not list it in `supportedSpeeds`). Omit it (the default) for
    * standard mode.