Sonnet 3.7 retired

Anthropic has shut off Claude Sonnet 3.7 and Claude Haiku 3.5 on its API, and requests to both models now fail. (platform.claude.com) The change appeared in Anthropic’s release notes on April 16, 2026, which say Sonnet 3.7 used the API ID `claude-3-7-sonnet-20250219` and Haiku 3.5 used `claude-3-5-haiku-20241022`. (platform.claude.com) Anthropic’s deprecation policy defines “retired” as no longer available for use, with requests to retired models failing instead of being silently rerouted. The company says customers with active deployments get at least 60 days’ notice before retirement for publicly released models. (platform.claude.com) The practical problem for developers is simple: model names are hard-coded into many apps, agents, and internal tools. If those strings still point to Sonnet 3.7 or Haiku 3.5, the calls now error out until the code is updated. (platform.claude.com) Anthropic is steering users to its newer 4.x line instead. Its current models page lists Claude Sonnet 4.6 as the “best combination of speed and intelligence” and Claude Haiku 4.5 as the fastest model in the lineup. (platform.claude.com) That recommendation lands in the middle of a faster release cycle. On April 14, 2026, Anthropic also said the original Claude Sonnet 4 and Claude Opus 4 models are deprecated and scheduled for retirement on June 15, 2026, with Sonnet 4.6 and Opus 4.7 as replacements. (platform.claude.com) The newer Sonnet 4.6 model is not a drop-in upgrade in every case. Anthropic’s migration docs say Sonnet 4.6 uses adaptive thinking as the recommended mode, while the older `thinking: {type: "enabled", budget_tokens: N}` approach is deprecated on 4.6 models and will be removed in a future release. (platform.claude.com) Anthropic also says Sonnet 4.6 supports a 1 million-token context window and 64,000 output tokens, while Haiku 4.5 is positioned for faster, cheaper, high-volume work at $1 per million input tokens and $5 per million output tokens. Sonnet 4.6 is priced at $3 per million input tokens and $15 per million output tokens. (platform.claude.com) For teams still on the retired models, Anthropic’s own advice is to export API usage from the Console, find which keys and services still call the old model IDs, and migrate them before the next retirement date hits. (platform.claude.com)