Anthropic is retiring Claude Opus 4.1 on August 5, 2026 — here's what developers need to do
Anthropic has deprecated claude-opus-4-1-20250805 with a retirement date of August 5, 2026. Opus 4.8 is the recommended replacement.
On June 5, Anthropic notified API developers that Claude Opus 4.1 (claude-opus-4-1-20250805) is now deprecated, with a firm retirement date of August 5, 2026. From that date, any requests to that model ID will fail. The recommended replacement is Claude Opus 4.8.
This applies to the Claude API, Claude Platform on AWS, and Microsoft Foundry. If you’re using this model through Amazon Bedrock or Vertex AI, those platforms set their own retirement schedules, so check with them directly.
What “deprecated” actually means
Anthropic uses four stages to describe a model’s lifecycle: Active, Legacy, Deprecated, and Retired.
Deprecated sits between legacy and gone. The model still works today, but Anthropic no longer recommends it, has assigned a retirement date, and will stop serving it entirely once that date arrives. You have roughly two months from the June 5 notification to migrate before things start breaking.
Anthropic’s policy is to give at least 60 days notice before retiring any publicly released model, so August 5 sits right at that minimum window. There’s no buffer built in beyond what they’ve already given you.
Why this is happening now
The Opus line has moved quickly. Claude Opus 4.1 launched in August 2025, and since then Anthropic has shipped Opus 4.5, 4.6, 4.7, and now 4.8. That’s a lot of model versions in a short span, and Anthropic has been explicit that it retires older models to free up capacity for newer ones.
Opus 4 (claude-opus-4-20250514) and Sonnet 4 (claude-sonnet-4-20250514) are also being retired, with a closer deadline of June 15, 2026. If you have any of those in production, those should be your first priority.
What Opus 4.8 brings to the table
Claude Opus 4.8 is Anthropic’s current most capable generally available model, and it’s a meaningful step up from 4.1 in a few specific ways.
The headline improvement is judgment. In agentic and coding contexts, Opus 4.8 is notably better at catching its own mistakes before they propagate, asking clarifying questions at the right moments, and pushing back on plans that don’t hold up. Anthropic’s own evaluations found it roughly four times less likely to leave flaws in its own code unremarked compared to its predecessor. If you’re using Opus in any kind of autonomous or multi-step workflow, that matters.
On context, Opus 4.8 supports a 1 million token context window on the Claude API, Amazon Bedrock, and Vertex AI. Microsoft Foundry gets 200k. Max output is 128k tokens.
Pricing starts at $5 per million input tokens and $25 per million output tokens. Prompt caching can reduce that by up to 90%, and batch processing by 50%. One small but useful detail: the minimum prompt length to qualify for caching is 1,024 tokens, lower than on Opus 4.7. Prompts that were previously too short to cache may now benefit without any code changes on your part.
What you need to do
If you’re hardcoding model IDs in your applications, this is the task: find every reference to claude-opus-4-1-20250805 and update it to claude-opus-4-8.
A few practical steps worth following:
Audit your usage now. Anthropic provides an API usage audit tool in the platform to help you locate deprecated model calls across your applications. Use it before you assume you’ve caught everything.
Test before the deadline. Don’t swap model IDs on August 4. Give yourself time to test prompts, check output quality, and validate that your application behaves as expected with Opus 4.8. The models are not identical, and depending on your use case, you may need to adjust prompts.
Note the model ID format. Opus 4.8’s ID is claude-opus-4-8 without a date suffix. Starting with the 4.6 generation, Anthropic moved to dateless model IDs that are still pinned snapshots, not evergreen pointers. You’re not getting automatic updates by using the shorter format.
Check your partner platforms separately. If any part of your stack runs through Bedrock or Vertex AI, the August 5 date doesn’t automatically apply. Those platforms will publish their own timelines.
If you don’t act
Requests to claude-opus-4-1-20250805 after August 5, 2026 will return errors. That’s not a soft failure or a graceful fallback to a newer model. The API call fails. For anything customer-facing or business-critical, that’s an outage.
Deprecated models are also likely to be less reliable than active ones in the months leading up to retirement, as Anthropic focuses engineering attention on current models. Early migration is worth it even before the deadline.
Two months is a comfortable window for most teams, but it goes quickly if it’s not on anyone’s sprint board. Worth getting it scheduled now.