Claude✓ ReviewedTest carefullyChecked 9 September 2026Claude Opus 5 can carry harder work. Old prompts may make it do too much.
Anthropic's newest Opus model expands long-context and agentic work, but teams should retest scope, effort and response length before replacing an older model in production.
Bit’s takeaway
What changed
Anthropic released Claude Opus 5 with a one million token context window, up to 128,000 output tokens, thinking enabled by default and standard API pricing of US$5 per million input tokens and US$25 per million output tokens.
Why it matters
A model migration is not only a model-ID change. Anthropic warns that Opus 5 can produce longer user-facing responses, narrate more of its work, widen an open-ended task and over-verify when old prompts contain repeated checking instructions.
Who should care
- Teams using Claude for complex coding or long-horizon work
- People moving a production workflow from an earlier Opus model
What to do
Run one representative task with the old and new model, keep the sources and acceptance criteria fixed, then compare scope drift, correction time, token use and the final reviewed result.
The human take
Tools change fast. Your judgment matters more.
The team still defines the boundary, reviews the result and decides whether the extra capability improved accepted work.
Affected guidance
Put this to work
Verified facts
- Anthropic released Claude Opus 5 with a one million token context window, up to 128,000 output tokens, thinking enabled by default and published standard API pricing.Checked
Sources and method