Claude Opus 4.1 retires on 5 August 2026: migrate to Opus 4.8 now
The claude-opus-4-1-20250805 model ID stops working on 5 August 2026. Here's what to change and why Opus 4.8 is the right target.
Topic
13 posts about Claude API from Simon Carter.
The claude-opus-4-1-20250805 model ID stops working on 5 August 2026. Here's what to change and why Opus 4.8 is the right target.
The legacy Workbench and three /v1/experimental/ prompt endpoints shut down 17 August 2026. Here's what to export and what replaces them.
Anthropic has deprecated claude-opus-4-1-20250805 with a retirement date of August 5, 2026. Opus 4.8 is the recommended replacement.
As of June 15, 2026, claude-sonnet-4-20250514 and claude-opus-4-20250514 return errors. Migrate to Sonnet 4.6 and Opus 4.8.
Two Claude outages in 14 hours on June 22–23 pushed 90-day uptime below enterprise SLA thresholds, with 8,000+ Downdetector reports at peak.
Anthropic has formally deprecated claude-mythos-preview with a June 30 2026 retirement date, directing developers to migrate to claude-mythos-5 — which is currently suspended.
From June 15, Claude Agent SDK, claude -p, GitHub Actions, and third-party agents draw from a separate monthly credit, not your subscription quota.
Anthropic launches Claude Fable 5 with a 1M-token context window, $10/$50 pricing, and a safety-classifier fallback — plus a restricted Mythos 5 for Project Glasswing partners.
Anthropic's vault-stored environment variables let Claude Managed Agents authenticate CLI tools without the API key ever entering the agent's context window.
Azure becomes the only cloud with both OpenAI and Claude frontier models, as Claude Opus 4.8 and Haiku 4.5 go GA in Microsoft Foundry on NVIDIA GB300 GPUs.
Anthropic ships three Claude Managed Agents platform updates giving enterprise developers more infrastructure control, runtime flexibility, and prompt-cache transparency.
Anthropic's new Swift package lets iOS/macOS 27 developers use Claude as a drop-in server-side model via Apple's LanguageModelSession API.
Anthropic now bills zero for API refusals with no output, and the advisor tool gets a max_tokens cap to control cost and latency in agentic workloads.