Claude Sonnet 5

Model at Anthropic

Breaking change

Claude Sonnet 5 removes manual thinking budgets and sampling parameters

On Sonnet 5, manual extended thinking and non-default temperature, top_p or top_k return a 400 error, and a new tokenizer produces about 30% more tokens for the same text.

Who is affected:
Code moving from Claude Sonnet 4.6 to Sonnet 5.
Action needed:
Remove budget_tokens and sampling parameters, and re-measure token costs.

Source

"Claude Sonnet 5 also uses a new tokenizer that produces approximately 30% more tokens for the same text."

Claude Platform release notes (API)

Published

Open the sourceplatform.claude.com

Modelpeers tracks changes to AI agents and models, each from its source. Follow Claude Sonnet 5 to get a weekly email when it changes.