Anthropic Claude Opus 4.8 and Dynamic Workflows

Model Updates and Performance

Anthropic released Claude Opus 4.8, which maintains the same pricing as its predecessor while offering improved judgment and speed. The model is approximately 2.5 times faster in 'fast mode', and the cost for this mode has been reduced to be three times cheaper than before, effectively costing 2x the standard rate for 2.5x the speed. Benchmark scores show a 5-point jump on SWE-bench Pro, reaching 69.2%, and notable improvements in multidisciplinary reasoning and agentic computer use. Despite these gains, GPT-5.5 remains the leader in terminal navigation benchmarks.

Dynamic Workflows

Anthropic introduced 'Dynamic Workflows' in Claude Code, a research preview feature designed for complex, long-horizon tasks like large-scale migrations or codebase-wide security audits. The system functions by having a main agent plan a task and then orchestrate hundreds of parallel sub-agents to execute subtasks. These sub-agents work independently and use adversarial prompting to refute findings before the main agent synthesizes a final result. This feature is accessible via the 'ultra code' setting in the Claude Code CLI, VS Code extension, and various API platforms. Anthropic also teased a future 'Mythos' class model, which is currently in limited preview for cybersecurity applications and expected to launch in the coming weeks.