AI models Retiring #330

Power Platform: Newer GPT models replace Anthropic, Grok, and earlier GPT models

Prompts move automatically to newer GPT models

Microsoft will automatically migrate affected prompts in Copilot Studio, Power Automate and Power Apps to newer OpenAI GPT models by December 2, 2026. Prompts will keep their current Basic, Standard, or Premium tier.

Key points

  • Basic: GPT-4.1 mini will be replaced by GPT-6 Luna.
  • Standard: GPT-4.1, GPT-5 chat, GPT-5.3 chat, Claude Sonnet 4.6, and Grok 4.1 Fast reasoning will be replaced by ChatGPT-latest.
  • Premium: GPT-5 reasoning, GPT-5.2 reasoning, and Claude Opus 4.6 will be replaced by GPT-6 Reasoning.
  • Tier pricing will not change, but actual consumption and costs can vary with usage and token counts.
  • Responses can change in wording, formatting, and how instructions are followed.

What to do

Identify affected prompts and their dependent apps, flows, and agents. Review and test them before December 2, 2026. Test each tier's replacement as it becomes available in the prompt builder, using representative inputs to check quality, output formats, and downstream processing. Adjust instructions or workflows where needed. Notify affected makers and support teams. Monitor results and consumption after migration. No action is required to trigger automatic migration.

Timeline

  1. 3 November 2026 Published
  2. 2 December 2026 Retirement end (planned) Listed Anthropic, Grok, and earlier GPT models will be fully retired.
  3. 2 December 2026 Rollout end (planned) Automatic migration of affected prompts will be complete.
Speaker notes

This change affects solutions that use prompts configured with the retiring models, rather than every solution in Power Platform. Microsoft will migrate those prompts automatically within their existing tier. Makers do not need to select a replacement, and retaining an old model selection will not prevent migration. The previous models will no longer be available in prompts after retirement. The main testing concern is whether changed responses still meet the output structures and behavior expected by dependent apps, flows, and agents. Tier pricing stays the same, but that does not guarantee unchanged costs because consumption can vary with usage and token counts.

Sources and links

Area, status, change type, licence and flags are indicative. Always open the source before you act on a card.