Tomorrow something happens that might catch quite a few developers off guard: Anthropic is shutting down Claude Sonnet 4 and Claude Opus 4 on the API. If you haven’t migrated by June 15, you’ll get error messages instead of responses.
What’s happening exactly
The original models from May 2025 — claude-sonnet-4-20250514 and claude-opus-4-20250514 — are being permanently retired. No grace period, no soft deprecation. API calls to these model strings will return errors starting June 15.
The migration itself is straightforward: update the model string, test your outputs, deploy. API format, authentication, and response structure remain identical. Anthropic recommends upgrading to Claude Sonnet 4.6 or Claude Opus 4.6 — and if you’re refactoring anyway, you might as well jump to Opus 4.7 for agentic workloads.
Agent SDK: Separate billing starts tomorrow
The second major change on the same day affects everyone who uses Claude programmatically. The Agent SDK, claude -p headless mode, GitHub Actions, and third-party apps built on the Agent SDK will no longer draw from your plan’s usage pool.
Instead, there’s a separate monthly credit: roughly $20 for Pro, $100 for Max 5x, and $200 for Max 20x, billed at standard API rates. Once that credit runs out, automated requests stop entirely — unless you’ve manually enabled overflow billing. No automatic fallback, no rollover.
Interactive usage — chat, Claude Code in the terminal, Cowork — is unaffected.
What this means
For teams running Claude in production pipelines, this is a real paradigm shift. If you’ve been running unlimited agent calls through your Max subscription, you now need to budget for it. Anthropic’s recommendation: for shared production automation, switch to Claude Platform with an API key for predictable pay-as-you-go billing.
The model retirement is routine — every provider does it regularly. But the Agent SDK split reveals where Anthropic is heading: agentic usage is becoming its own measurable business segment, separate from consumer subscriptions.
Sources: