Add Claude Fable 5 model

[juliusmarminge]

Summary

• add claude-fable-5 to the Claude provider built-in model catalog
• expose it as Claude Fable 5 with the provider default capabilities
• cover model availability in the Claude provider status test

Validation

• vp test apps/server/src/provider/Layers/ProviderRegistry.test.ts
• vp check
• vp run typecheck

Note

Add Claude Fable 5 model gated behind Claude Code version 2.1.169

• Adds a new claude-fable-5 built-in model with configurable reasoning (low/medium/high/xhigh/max/ultracode/ultrathink) and context window (200k/1M) options.
• Gates the model behind a supportsClaudeFable5 version check in ClaudeProvider.ts; the model only appears when the detected Claude Code version is >= 2.1.169.
• Adds a formatClaudeFable5UpgradeMessage helper and updates checkClaudeProviderStatus to show a Fable 5-specific upgrade prompt when the version is insufficient.
• Adds a Fast Mode boolean capability to the existing claude-opus-4-8 and claude-opus-4-7 models.

Changes since #3009 opened

• Modified normalizeClaudeCliEffort function to preserve xhigh effort option for claude-fable-5 model [9f57633]
• Refactored model capability descriptors in BUILT_IN_MODELS constant [9f57633]
• Added test coverage for claude-fable-5 model with xhigh effort option [9f57633]
• Refactored effort options configuration for Claude models [cbe8a20]

^{Macroscope summarized e8d43cb.}

---

Note

Low Risk
Provider catalog and effort normalization only; no auth or data-path changes, covered by unit tests.

Overview
Adds Claude Fable 5 (claude-fable-5) to the Claude built-in catalog with reasoning and context-window options, and only exposes it when Claude Code is ≥ 2.1.169; older installs get a Fable-specific upgrade message in provider status.

Effort handling treats Fable 5 like Opus 4.8 for xhigh (no remap to max). Shared CLAUDE_EFFORT_OPTIONS is removed in favor of per-model descriptors; Fast Mode is added on Opus 4.8 and 4.7. Tests cover Fable 5 availability by CLI version and adapter xhigh passthrough.

^{Reviewed by Cursor Bugbot for commit cbe8a20. Bugbot is set up for automated code reviews on this repo. Configure here.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: cc857648-4ecc-4d29-868f-1c0f3bccd6c3

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/add-claude-fable-5

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

🚀 Expo continuous deployment is ready!

• Project → t3-code
• Platforms → android, ios
• Scheme → t3code-preview

	🤖 Android	🍎 iOS
Fingerprint	ae17d94b35f91f9c608a63dadbc3ddd9f4ba056e	313e506a7f15a9c3f15b01d787d68a12382432bf
Build Details	Build Permalink Details Distribution: INTERNAL Build profile: preview:dev Runtime version: ae17d94b35f91f9c608a63dadbc3ddd9f4ba056e App version: 0.1.0 Git commit: 192f84f2bb2814a4b21ac59218fc7e1d5c4958a4	Build Permalink Details Distribution: INTERNAL Build profile: preview:dev Runtime version: 313e506a7f15a9c3f15b01d787d68a12382432bf App version: 0.1.0 Git commit: f23a3711d86cd8b6f514f05926830d7f83f9f5c1
Update Details	Update Permalink Details Branch: pr-3009 Runtime version: ae17d94b35f91f9c608a63dadbc3ddd9f4ba056e Git commit: 192f84f2bb2814a4b21ac59218fc7e1d5c4958a4	Update Permalink Details Branch: pr-3009 Runtime version: 313e506a7f15a9c3f15b01d787d68a12382432bf Git commit: 192f84f2bb2814a4b21ac59218fc7e1d5c4958a4
Update QR

[macroscopeapp]

Approvability

Verdict: Approved

Straightforward addition of a new model to the provider registry following existing patterns exactly, with comprehensive tests. The refactoring inlines a shared constant but values remain identical. The open review comment is a minor naming question, not a correctness issue.

^{You can customize Macroscope's approvability policy. Learn more.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Fable xhigh effort downgraded incorrectly
  • Added claude-fable-5 to the exemption check in normalizeClaudeCliEffort so xhigh is no longer incorrectly downgraded to max for that model.

Or push these changes by commenting:

@cursor push 7ec07d5b83

Preview (7ec07d5b83)

diff --git a/apps/server/src/provider/Layers/ClaudeProvider.ts b/apps/server/src/provider/Layers/ClaudeProvider.ts
--- a/apps/server/src/provider/Layers/ClaudeProvider.ts
+++ b/apps/server/src/provider/Layers/ClaudeProvider.ts
@@ -334,7 +334,7 @@
   if (effort === "ultracode") {
     return "xhigh";
   }
-  if (effort === "xhigh" && model !== "claude-opus-4-8") {
+  if (effort === "xhigh" && model !== "claude-opus-4-8" && model !== "claude-fable-5") {
     return "max";
   }
   if (effort === "max" && model === "claude-sonnet-4-6") {

_{You can send follow-ups to the cloud agent here.}

^{Reviewed by Cursor Bugbot for commit 648538a. Configure here.}

[buzinas]

Shouldn't this be fable5 (or maybe fable50)?

[cursor]

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Fable capabilities contradict test
  • Updated the test assertion to verify the actual effort and contextWindow option descriptors defined on the claude-fable-5 model, instead of incorrectly expecting empty optionDescriptors.

Or push these changes by commenting:

@cursor push 539a67e334

Preview (539a67e334)

diff --git a/apps/server/src/provider/Layers/ProviderRegistry.test.ts b/apps/server/src/provider/Layers/ProviderRegistry.test.ts
--- a/apps/server/src/provider/Layers/ProviderRegistry.test.ts
+++ b/apps/server/src/provider/Layers/ProviderRegistry.test.ts
@@ -1350,15 +1350,14 @@
           assert.strictEqual(status.status, "ready");
           assert.strictEqual(status.installed, true);
           assert.strictEqual(status.auth.status, "authenticated");
-          assert.deepStrictEqual(
-            status.models.find((model) => model.slug === "claude-fable-5"),
-            {
-              slug: "claude-fable-5",
-              name: "Claude Fable 5",
-              isCustom: false,
-              capabilities: { optionDescriptors: [] },
-            },
-          );
+          const fableModel = status.models.find((model) => model.slug === "claude-fable-5");
+          assert.strictEqual(fableModel?.slug, "claude-fable-5");
+          assert.strictEqual(fableModel?.name, "Claude Fable 5");
+          assert.strictEqual(fableModel?.isCustom, false);
+          const descriptors = fableModel?.capabilities?.optionDescriptors ?? [];
+          assert.strictEqual(descriptors.length, 2);
+          assert.strictEqual(descriptors[0]?.id, "effort");
+          assert.strictEqual(descriptors[1]?.id, "contextWindow");
         }).pipe(
           Effect.provide(
             mockSpawnerLayer((args) => {

_{You can send follow-ups to the cloud agent here.}