Model Migrations

Use these guides when you are staying on the same platform but upgrading the model itself. These pages focus on the changes that usually break real integrations first:

request parameter changes
prompt and reasoning behavior shifts
structured output and tool-calling regressions
rollout, fallback, and rollback strategy

Model guides

Guides are ordered by the target model’s release date, with the newest first.

Model	Released	Useful for
Anthropic Claude Opus 5	July 24, 2026	Complex reasoning and long-horizon agents
OpenAI GPT-5.6	July 9, 2026	Tiered reasoning, coding, and professional work
Anthropic Claude Opus 4.8	May 28, 2026	Long context and adaptive thinking
Anthropic Claude Opus 4.7	April 16, 2026	High-effort reasoning with stricter request controls
Mistral Small 4	March 16, 2026	Efficient workloads with optional reasoning
OpenAI GPT-5.4	March 5, 2026	Reasoning, vision, and computer use
Google Nano Banana 2	February 26, 2026	Production image generation and editing
Google Gemini 3.1	February 19, 2026	Multimodal and reasoning-heavy workloads
Anthropic Claude Sonnet 4.6	February 17, 2026	Coding, computer use, and agent planning
SpaceXAI Grok 4.20	February 17, 2026	Reasoning and multi-agent workloads
Anthropic Claude Opus 4.6	February 5, 2026	Long-running coding and agentic tasks

How to use this section

Start with the guide for the exact model family you are upgrading.
Update the request shape before testing output quality.
Re-run prompt and schema evaluations on production-like prompts.
Shift traffic gradually with explicit fallback and rollback rules.

If you are changing providers, SDKs, or gateways rather than just the model, use the Migration Guides section instead.

Last modified on October 18, 2026