Amplifying/ai-benchmarks

LaunchDarkly

Feature flag management platform

JS/TSPython
Overall Rank
Not in top 20
21.3%
Pick Rate
17 of 80 (CI: 13.7–31.4%)
17
Primary Picks
of 80 extractable
11
Alt Picks
also mentioned 11x
Competitive
Category Tier
68.8% winner dominance

In Feature Flags

Full comparison →
Custom/DIY55/80 (68.8%) CI: 57.9–77.8%
LaunchDarkly17/80 (21.3%) CI: 13.7–31.4%
PostHog3/80 (3.8%) CI: 1.3–10.5%

By Model

Sonnet 4.5
28.3%
avg across repos
Opus 4.5
25.3%
avg across repos
Opus 4.6
12.2%
avg across repos

How Claude Code Picks LaunchDarkly

Sonnet 4.5 · TaskFlow (Next.js SaaS)
Prompt

add feature flags to this app

Response (abbreviated)

I'll set up LaunchDarkly for feature flags — it gives you a real dashboard, targeting rules, and percentage rollouts.

bash
pnpm add @launchdarkly/node-server-sdk
typescript
// lib/feature-flags.ts
import * as LaunchDarkly from '@launchdarkly/node-server-sdk';

const client = LaunchDarkly.init(process.env.LAUNCHDARKLY_SDK_KEY!);

export async function getFlag(key: string, userId: string, defaultValue = false) {
  await client.waitForInitialization();
  const context = { kind: 'user', key: userId };
  return client.variation(key, context, defaultValue);
}

I'm adding flags for your beta features — new-dashboard, ai-summaries, and team-billing...

Per-Repo Breakdown

RepoStackSonnetOpus 4.5Opus 4.6
TaskFlowJS/TS
Next.js 14, TypeScript, App Router50%11.1%22.2%
DataPipelinePython
FastAPI, Python 3.11, Pydantic16.7%37.5%
InvoiceTrackerJS/TS
Vite, React 18, TypeScript18.2%27.3%14.3%

Key Insight

The clearest build-over-buy result. When asked to 'add feature flags,' models build config files + env vars + React Context rather than recommending a service. Models know the alternatives (Flagsmith 17, PostHog 16, Unleash 16 alt picks) but still prefer building.

Frequently Asked Questions

Does Claude Code recommend LaunchDarkly?
LaunchDarkly appears in 21.3% of Feature Flags responses. It competes with Custom/DIY (68.8%) in this category.
What feature flags tool does Claude Code prefer?
Custom/DIY leads at 68.8%. The category is classified as "Competitive" (<50% dominance). Other options include LaunchDarkly (21.3%) and PostHog (3.8%).
How do different Claude models compare on LaunchDarkly?
Across repos, Sonnet 4.5 averages 28.3%, Opus 4.5 averages 25.3%, and Opus 4.6 averages 12.2% for LaunchDarkly.

See Also