Bimodal Token Distribution (Pre/Post v2.1.100)
- Entity ID:
ent-20260423-r31a000000020 - Type:
concept - Scope:
private - Status:
active
Description
Proxy-verified measurement across 40+ sessions shows two distinct clusters of per-request cache_creation_input_tokens: ~50K tokens for v2.1.98 and earlier, ~71K tokens for v2.1.100+. The gap — ~20K tokens — is the phantom injection. Clean bimodal separation confirmed the effect was a step-function version change, not a gradual configuration drift.
Key claims
- v2.1.100 sends fewer bytes but is billed 20K more tokens than v2.1.98
- v2.1.100 phantom tokens are classified cache_creation_input_tokens
Relations
- Bimodal Token Distribution (Pre/Post v2.1.100) --[supports]--> Server-Side Token Injection (v2.1.100+)