tech
Claude's New Model Is More 'Honest' When It Messes Up
Posts from this topic will be added to your daily email digest and your homepage feed.
TL;DR
- Anthropic is releasing Claude Opus 4.8, focusing on 'honesty' in AI responses.
- The new model is trained to avoid unsupported claims and flag uncertainties.
- Early testers report Opus 4.8 is more likely to admit when it doesn't know something.
- It is approximately four times less likely than its predecessor to overlook flaws in generated code.
- Users can now direct the amount of effort Claude expends on a task, balancing response quality with token usage.
- A new 'dynamic workflows' feature allows Claude to plan and execute hundreds of parallel subagents for larger tasks.
- With dynamic workflows, Claude can verify outputs before reporting back to the user.