xhigh Effort Level
- Entity ID:
ent-20260423-r30a000000010 - Type:
concept - Scope:
private - Status:
active
Description
New Opus 4.7 effort tier that sits above high in the effort hierarchy. Allocates a larger thinking budget (more internal tokens) and is tuned for longer, more globally coherent reasoning. Available to Max subscribers in auto mode. A typical coding turn at high consumes 5-10k tokens; xhigh can double that. Combined with the 1.35x tokenizer overhead, a heavy auto-mode xhigh session with poor caching can be 2-3x more expensive than the same workflow on 4.6 high.
Key claims
- xhigh sits above high in the Opus 4.7 effort hierarchy
- /effort without arguments now opens an interactive slider
- xhigh + 1.35x tokenizer + bad cache = 2-3x cost vs 4.6 high
- Use Opus 4.7 high for most coding; reserve xhigh for hard reasoning
Relations
- xhigh Effort Level --[contains]--> Claude Opus 4.7
- /effort Interactive Slider --[depends_on]--> xhigh Effort Level
- Tokenizer-Effort-Cache Cost-Multiplier Model --[depends_on]--> xhigh Effort Level
- Post-Leak Claude Code Playbook --[depends_on]--> xhigh Effort Level