Claude 4 Deprecation: Complete Migration Action Plan
Claude 4 Opus and Sonnet 4 were retired on June 15, 2026. After that date, every API call to these models fail. Here's exactly what to do, day by day, to migrate safely — and save 67-99% in the process.
claude-4-opus and claude-sonnet-4 return errors. No grace period. No extensions. Migrate now.
TL;DR: The Minimum Viable Migration
If you just need to keep your app running, here's the 5-minute fix:
- Find where your code specifies
claude-4-opusorclaude-sonnet-4 - Replace with
claude-opus-4-8orclaude-sonnet-4-6 - Deploy
That's it. Your API key, SDK, and endpoint stay the same. You'll also save 67% on Opus costs.
The 10-Day Action Plan
Here's what to do each day from now until the deadline. Adjust based on your project complexity.
Audit: Find Every Reference to Claude 4
Search your entire codebase for claude-4-opus and claude-sonnet-4. Check these locations:
- Source code (API calls, model configuration)
- Environment variables and config files
- Documentation and README files
- CI/CD pipelines and deployment scripts
- Monitoring and logging setup (model name in dashboards)
- Third-party integrations (LangChain, LlamaIndex, etc.)
Command: grep -r "claude-4-opus\|claude-sonnet-4" .
Choose Your Replacement
You have two paths:
Path A: Stay with Anthropic (easiest)
claude-4-opus→claude-opus-4-8(67% cheaper, 1M context)claude-sonnet-4→claude-sonnet-4-6(same price, 1M context)
Path B: Switch providers (bigger savings)
- GPT-5: $1.25/$10 — great all-around
- Gemini 3.1 Pro: $2/$12 — best value at scale
- DeepSeek V4 Pro: $0.435/$0.87 — cheapest by far
Use the cost calculator to model your actual usage.
Make the Code Changes
Update model name strings in your codebase. For Anthropic, only the model name changes:
- Python:
model="claude-opus-4-8" - Node.js:
model: "claude-opus-4-8" - cURL:
"model": "claude-opus-4-8"
If switching providers, also update the API endpoint and install the new SDK.
Test in Staging
- Run your test suite against the new model
- Verify API responses match expected format
- Test edge cases: long prompts, streaming, function calling
- Check token usage (new models have different tokenizers)
Update Dependencies & Config
- Update LangChain, LlamaIndex, or other frameworks to latest versions
- Check if your framework has a
ChatAnthropicor similar wrapper that needs updating - Update any hardcoded model lists in dashboards or admin panels
Stress Test & Benchmark
- Run a production-like load test
- Compare latency: old vs new model
- Monitor error rates for the first hour
- Verify rate limits haven't changed
Deploy to Production
- Deploy the migration to production
- Keep the old model config as a commented-out fallback
- Monitor logs for the first 24 hours
- Set up alerts for error rate spikes
Update Documentation
- Update README with new model name
- Notify team members about the change
- Update any client-facing documentation
Final Verification
- Double-check no remaining references to
claude-4-opusorclaude-sonnet-4 - Verify production is stable
- Review cost savings in your billing dashboard
Rest Easy
You're migrated. Your app is running on a cheaper, faster model with a bigger context window. The June 15 deadline won't affect you.
Cost Comparison: Before vs After
Here's what the migration actually costs you (spoiler: it saves money):
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Context Window | Monthly Cost* |
|---|---|---|---|---|
| Claude 4 Opus (deprecated) | $15.00 | $75.00 | 200K | $300.00 |
| Claude Opus 4.8 | $5.00 | $25.00 | 1M | $100.00 (save 67%) |
| GPT-5 | $1.25 | $10.00 | 272K | $32.50 (save 89%) |
| Gemini 3.1 Pro | $2.00 | $12.00 | 1M | $44.00 (save 85%) |
| DeepSeek V4 Pro | $0.435 | $0.87 | 1M | $6.09 (save 98%) |
*Based on 10M input + 2M output tokens/month. Calculate your actual cost →
What If You're Using Claude Sonnet 4?
Sonnet 4 → Sonnet 4.6 is a same-price swap with a 5x larger context window:
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Context Window |
|---|---|---|---|
| Claude Sonnet 4 (deprecated) | $3.00 | $15.00 | 200K |
| Claude Sonnet 4.6 | $3.00 | $15.00 | 1M |
Same price, 5x more context. This is a no-brainer upgrade.
Common Migration Pitfalls
- Hardcoded model names in config files. Check
.env,config.yaml, anddocker-compose.yml. - Third-party framework defaults. LangChain, LlamaIndex, and other frameworks may have their own model name mappings.
- Batch jobs and cron tasks. These often use separate config and get forgotten.
- Multiple API keys. If you have separate keys for different environments, update all of them.
- Model name in logs and dashboards. Your monitoring might break if it filters by model name.
Don't Just Migrate — Optimize
The deprecation is an opportunity to cut costs by 67-98%. Use our free tools to find the cheapest model for your workload.
Calculate Your Savings →Related Resources
🎯 Rate Your API Setup in 30 Seconds
Get an A+ to F grade on your AI API costs. See how you compare and find cheaper alternatives instantly.
Get Your Cost Score →📊 Generate Your Personalized API Cost Report
Select your model, enter your monthly spend, and get a custom savings report with cheaper alternatives — free, in 60 seconds.
Want to optimize your AI API costs?
APIpulse includes free cost comparisons, exports, and recommendations that can save you up to 40%.
Free Cost Audit →