Market Briefs / technology
AI model shutdowns hit client apps on November 30 and December 1
Three AI shutdowns land within 24 hours. Anthropic retires Claude Sonnet 4.5 on November 30, OpenAI shuts down Agent Builder and its Evals platform the same day, and from December 1 Google routes gemini-2.5-pro and gemini-2.5-flash calls to newer models billed at their rates.
Published
What happened?
Anthropic deprecated claude-sonnet-4-5-20250929 on September 30 and retires it on the Claude API on November 30, 2026. Requests to it fail after that date, and Anthropic names Claude Sonnet 5.5 as the replacement; Amazon Bedrock and Google-managed Claude on Google Cloud set their own dates. OpenAI’s Agent Builder and the Evals dashboard and API are scheduled to shut down on November 30; existing evals became read-only on October 31, and ChatKit remains available. On November 2 Google deprecated gemini-2.5-pro and gemini-2.5-flash: from December 1, requests to them are automatically routed to gemini-3.8-flash and gemini-3.5-flash-lite respectively and billed at those models’ rates. Google lists breaking changes including function calling and thought_signature validation, thinking_level replacing thinking_budget, and removed sampling parameters. gemini-2.5-flash-lite is not affected.
Why it matters
These failures tend to surface as a client message, not a deploy error. Claude calls fail outright. Google’s reroute keeps answering, but with a different model, different output, possibly broken tool calls and a different bill. Freelancers who built chatbots, automations or agent workflows in 2025 are the most likely to have these model IDs hard-coded in projects they no longer actively maintain.
What should you do?
Search every client repository, environment file and automation scenario (n8n, Make, Zapier) for “claude-sonnet-4-5”, “gemini-2.5-pro”, “gemini-2.5-flash” and Agent Builder workflow IDs. Tell affected clients now, with the date, and quote the migration as scoped work: change the model ID, apply the provider’s migration guide, re-test prompts and tool calls, and compare cost. Export any Evals data before November 30 and move workflows to the Agents SDK or an evaluation tool such as Promptfoo.