OpenAI retires GPT-4, GPT-4-turbo and the o-series on Oct 23
A model-retirement wave hits this month: Alibaba cuts legacy models Oct 10, Azure drops Kimi today, and pinned API calls fail unless you migrate first.
Copy markdown
OpenAI pulls GPT-4 and the o-series Oct 23
On Oct 23 OpenAI shuts down gpt-3.5-turbo, gpt-4, gpt-4-0613, gpt-4-turbo, o1, o1-pro, o3-mini, gpt-4.1-nano and gpt-image-1 - plus any fine-tunes built on them. Calls redirect to gpt-5.6-sol, terra, or luna (and gpt-image-2.5), but outputs and latency shift, so re-run your prompts and evals before the cutoff instead of trusting a silent swap.
Alibaba Model Studio purges legacy models Oct 10
Alibaba Cloud Model Studio retires a batch of legacy snapshot models on Oct 10 (UTC+08); after that, calls to them return nothing at all. Deprecation trackers put the scope near 14 families - DeepSeek r1/v3/v3.x, GLM, Kimi, MiniMax and older Qwen3 - folding onto qwen3.7-plus and qwen3.6-flash. If your stack routes through Model Studio, migrate and re-validate this week.
Also going dark: Kimi on Azure, Sonnet 4 on Bedrock
Per deprecation trackers, Moonshot's Kimi leaves Azure AI Foundry today with no named successor, and Anthropic's claude-sonnet-4-20250514 exits Amazon Bedrock on Oct 14 - both hand you the job of choosing a replacement. The pattern is clear: pin a model and a cloud will eventually pull it, so keep an abstraction layer and a fallback route.