fal-optimization

## Quick Reference

| Optimization | Technique | Impact |
|--------------|-----------|--------|
| Parallel requests | `Promise.all()` with batches | 5-10x throughput |
| Avoid polling | Use webhooks | Lower API calls |
| Cache by seed | Store `prompt+seed` results | Avoid regeneration |
| Right-size images | Use needed resolution | Lower cost |
| Fewer steps | Reduce inference steps | Faster, cheaper |

| Model Tier | Development | Production |
|------------|-------------|------------|
| Image | FLUX Schnell | FLUX.2 Pro |
| Video | Runway Turbo | Kling 2.6 Pro |

| Serverless Config | Cost-Optimized | Latency-Optimized |
|-------------------|----------------|-------------------|
| `min_concurrency` | `0` | `1+` |
| `keep_alive` | `120` | `600+` |
| `machine_type` | Smallest viable | Higher tier |

## When to Use This Skill

Use for **performance and cost optimization**:
- Reducing generation latency
- Lowering API costs
- Implementing parallel processing
- Choosing between polling and webhooks
- Configuring serverless scaling

**Related skills:**
- For API patterns: see `fal-api-reference`
- For model selection: see `fal-model-guide`
- For serverless config: see `fal-serverless-guide`

---

# fal.ai Performance and Cost Optimization

Strategies for optimizing performance, reducing costs, and scaling fal.ai integrations.

## Performance Optimization

### Client-Side Optimizations

#### 1. Use Queue-Based Execution

Always prefer `subscribe()` over `run()` for generation tasks:

```typescript
// Recommended: Queue-based with progress tracking
const result = await fal.subscribe("fal-ai/flux/dev", {
  input: { prompt: "test" },
  logs: true,
  onQueueUpdate: (update) => {
    // Show progress to users
    if (update.status === "IN_PROGRESS") {
      console.log("Generating...");
    }
  }
});

// Only use run() for fast endpoints (< 30s)
const quickResult = await fal.run("fal-ai/fast-sdxl", {
  input: { prompt: "quick test" }
});
```

#### 2. Parallel Requests

Process