Claude Opus 4.1 is retired: what breaks on 5 August 2026 and how to fix it
The claude-opus-4-1-20250805 model ID retired on 5 August 2026. API calls to it now return a hard 404. Here is what to do.
As of 5 August 2026, the model ID claude-opus-4-1-20250805 is retired on Anthropic’s platform. There is no graceful fallback, no redirect to a newer model, and no warning in the response body. Requests to that endpoint return a hard 404 error. If your production code pins that model ID, it is broken right now.
What happened and when
Anthropic announced the deprecation of claude-opus-4-1-20250805 on 5 June 2026, giving teams a 60-day window to migrate. During that window the model continued to work, but Anthropic flagged it as no longer recommended. The retirement date of 5 August 2026 was the hard stop: from that point, the endpoint returns errors and there is nothing to catch them on Anthropic’s side.
This follows the same pattern Anthropic used for earlier retirements. Claude 3 Opus went on 5 January 2026. Claude Sonnet 4 and the original Claude Opus 4 retired together on 15 June 2026. Anthropic is now running a deprecation cadence of roughly 60 to 90 days after a successor ships, so if you have not already built migration into your maintenance routine, this is a good moment to start.
What breaks, exactly
The retirement applies to Anthropic’s own platforms: the Claude API, Claude Platform on AWS, and Microsoft Foundry. If you are calling claude-opus-4-1-20250805 directly through any of those, every request fails from 5 August 2026 onwards.
Partner-operated platforms, Amazon Bedrock and Google Cloud Vertex AI, set their own retirement schedules independently. If you are routing through one of those, check their documentation separately. The date above does not apply to them.
The failure mode is worth spelling out clearly. There is no automatic fallback to a newer model. Anthropic does not silently redirect the request. The dated model ID claude-opus-4-1-20250805 simply returns an error, full stop. Any retry logic that keeps hitting the same endpoint will keep failing.
What to migrate to
Anthropic’s recommended replacement is Claude Opus 4.8. For complex coding tasks, multi-step reasoning, and agentic workflows, Opus 4.8 is the current active model in the Opus tier and the sensible default for anything that was previously running on Opus 4.1.
One technical note worth flagging: if you migrate to Opus 4.7 or 4.8, the temperature, top_p, and top_k parameters now return an HTTP 400 error if you pass non-default values. You need to remove those fields from your requests and steer model behaviour through prompting instead. This is not a minor config tweak if your existing code relies heavily on temperature tuning, so factor in testing time.
Also worth noting: there is no public Opus 4.2, 4.3, or 4.4. Anthropic moved from 4.1 directly to the 4.5 and above lineage. Do not go looking for intermediate versions.
What this means for your codebase
Run a grep across your codebase, infrastructure configs, and environment variables for every string matching claude-opus-4-1. Check:
- Direct API call parameters
- SDK initialisation defaults
- LangChain or other wrapper configurations
- Routing configs and load balancer rules
- Monitoring dashboards and alert definitions that might reference the model ID
- Any CI/CD pipelines that call the API as part of testing
If you use dateless alias IDs rather than pinned dated IDs, you may already be pointing at a current model automatically. Check Anthropic’s model IDs documentation to confirm what your alias resolves to.
Once you have updated your model ID, test your fallback chain before you consider the migration complete. Set up a synthetic request that deliberately triggers each fallback tier in your routing layer and verify each one returns a 200 before you sign off.
Building a better habit going forward
This retirement will not be the last. Anthropic’s stated policy is to give at least 60 days’ notice before retiring any publicly released model, and they notify customers with active deployments by email as well as updating the deprecations page. That page is the canonical source and worth bookmarking.
The practical lesson is that pinning to dated model IDs gives you stability and predictability within a version’s lifetime, but it creates a hard cliff at retirement. Factoring a model audit into your team’s regular maintenance cycle, triggered on every Anthropic deprecation announcement, is a much cheaper habit than firefighting a production outage.
If you are not already subscribed to Anthropic’s deprecation notifications, check your account settings on the Claude Platform. The email notifications are the earliest official signal you will get that a model you depend on is entering its deprecation window.