Q: How does credit consumption work for Claude 3.5 Sonnet regarding input/output length and response complexity?
I understand Claude 3.5 Sonnet costs 7 credits per message. I need clarification on whether credit consumption remains fixed regardless of:
Input prompt length
Output response length
Task complexity
Code blocks/programming outputs
Multiple revisions within same response
Special formatting (tables, lists, markdown)
Mathematical equations/scientific notations
Long responses split into parts
Specific scenarios I'd like clarity on:
-Would a one-line response cost the same as a multi-page analysis?
-Are additional credits charged for revising/expanding initial responses?
-Do split responses due to length require extra credits?
-Is there a maximum length limit per message before additional credits apply?
-Does complex formatting affect credit consumption?
This information will help me better plan my credit usage and subscription needs.

Kevin_Zemith
Edited Feb 1, 2025A: Hi there,
Credits are consumed on a per-invocation basis. For chat, this means each assistant message response incurs a cost, regardless of complexity—whether it's a table, code block, or general conversation.
We recently introduced a higher message limit. For user messages exceeding 50,000 characters, the credit cost will be doubled. Otherwise, credits are charged per invocation as usual.