⚡️⚡️⚡️ Anthropic released Claude Opus 4.8.
The main highlight is that the model is four times less likely to miss bugs in the code it writes itself. For agent tasks, this is more important than another benchmark improvement.
The prices remain the same as 4.7 - $5/$25 per million tokens. And the fast mode has become three times cheaper, now costing $10/$50.
New features include: Dynamic Workflows (research preview) - Claude can orchestrate hundreds of parallel subagents for large codebase tasks, and Effort Control to adjust the model's "effort" from minimum to maximum.