If you’ve been relying on the "old reliable" versions of Claude to handle your daily coding or writing tasks, you probably woke up to a bit of a shock. Honestly, it’s the end of an era. On October 28, 2025, Anthropic officially pulled the plug on the Claude 3.5 Sonnet legacy models, forcing a massive migration that has developers scrambling and power users relearning their favorite workflows.
Basically, the "Computer Use" pioneer we all grew to love—the version of Sonnet that first learned how to move a cursor like a human back in late 2024—is officially retired. If your API calls started throwing errors today, this anthropic news october 28 2025 is exactly why.
The Great Model Migration: What Just Happened?
Anthropic doesn't usually move this fast, but the jump to the 4.5 architecture has been aggressive. As of today, any requests sent to the original claude-3-5-sonnet-20240620 or the upgraded claude-3-5-sonnet-20241022 (the one that introduced the famous Computer Use beta) will simply return an error.
You've probably seen the warnings in the developer console for months. But let’s be real, most of us wait until the last second to update our dependencies.
What was retired today:
- Claude 3.5 Sonnet (Original): The June 2024 model that first put GPT-4o on notice.
- Claude 3.5 Sonnet (V2): The October 2024 "Computer Use" trailblazer.
- Claude 3.7 Sonnet: Interestingly, even the 3.7 "hybrid reasoning" model is being deprecated in favor of the newer 4.5 logic.
The push is clearly toward Claude Sonnet 4.5. Anthropic claims it's better at basically everything, but if you’re a developer who spent weeks prompt-engineering for 3.5's specific quirks, today is kinda painful.
Why the October 28 Deadline Matters for Developers
This isn't just a routine update. Anthropic is fundamentally changing how it manages "thinking blocks." Along with the model retirements, they've introduced a feature called clear_thinking_20251015.
If you’re working with the newer reasoning models, you know they can get "chatty" in the background while they figure out a problem. This new expansion allows for better context editing. It basically lets the system automatically clear out those "internal monologue" blocks so your context window doesn't get cluttered with the AI's "brainstorming" sessions. It saves tokens, and honestly, it saves money.
The shift to platform.claude.com
If you tried to go to console.anthropic.com this morning, you probably noticed a redirect. The brand consolidation is almost complete. Anthropic is moving everything under the "Claude" umbrella. While the old URL still works for now, it’s officially on life support until January.
The $1.5 Billion Elephant in the Room
We can't talk about anthropic news october 28 2025 without mentioning the legal drama that’s finally coming to a head this month. While the tech updates are exciting, Anthropic is currently navigating the fallout of that massive $1.5 billion settlement with book authors.
Judge William Alsup has been keeping a very close watch on how Anthropic handles the "pirated" data from those 500,000 books. Part of the agreement involves not just paying authors about $3,000 per book, but also destroying the specific training data derived from those pirated sources.
Some industry insiders think the sudden retirement of the 3.5 family might be partially linked to this. If those older models were trained on datasets that are now legally "toxic," moving the entire user base to the 4.5 architecture—which was presumably trained with more rigorous, licensed data—is a smart legal move.
How Claude 4.5 Changes Your Workflow
If you’re moving to Sonnet 4.5 today because your old 3.5 keys stopped working, here’s what you’re actually getting. It's not just a speed boost.
- Agentic Reliability: The "Computer Use" feature is no longer a clunky beta. In the 4.5 version, it can handle multi-step procurement workflows (like actually buying something on a website) with about a 30% higher success rate than the models that retired today.
- Token Efficiency: With the new prompt caching and thinking block management, you aren't paying as much for repetitive instructions.
- Better "Vibe" Check: One of the big complaints about the newer models was that they felt a bit more "robotic." Anthropic has supposedly tuned 4.5 to regain some of that human-like warmth that made the original 3.5 Sonnet so popular.
Actionable Steps: What You Need to Do Right Now
If your apps are broken or you're just confused by the new interface, here's the game plan:
- Update your Model IDs: Switch your API calls from
claude-3-5-sonnet-20241022toclaude-sonnet-4-5-20250929. It’s a drop-in replacement, but you’ll want to test your temperature settings. - Audit your Context: Check if you can implement the
clear_thinkingblock. If you’re running long conversations, this will slash your token usage significantly. - Check your Privacy Settings: Remember that Anthropic updated their terms earlier this month. If you haven't opted out, your data might be used for training the upcoming Opus 4.5. You have until the end of this month to lock those settings down if you're on a consumer plan.
- Move your Bookmarks: Start using
platform.claude.comfor the developer dashboard. The old console is a ghost town now.
The transition today marks a massive shift in the AI landscape. We're moving away from the "experimental" phase of 2024 and into a world where AI models are expected to be autonomous agents. It’s a bit of a bumpy ride, but the capabilities of the 4.5 family are honestly on another level compared to the models we just said goodbye to.
🔗 Read more: How to Delete Devices from Hulu So Your Account Actually Stays Private
Update your API keys, check your prompts for "thinking" clutter, and make sure your team is ready for the 4.5 era.