I enabled Artifacts in Claude 3.7 Sonnet and attempted to output HTML, but the output was only displayed partially.

I measured the token count, and character count, of the input prompt and Artifacts output.

Input prompt (including Japanese):
- Token count: 7587 tokens (counted with Anthropic API)
- Character count: 8614 characters

Artifacts output (incomplete HTML including Japanese):
- Token count: 4065 tokens (counted with Anthropic API)
- Character count: 8475 characters

The output token count is consistently around 4K.

The FAQ mentions "Long Messages (>50K Chars): 2x model cost"

What does this FAQ entry refer to? Based on my measurements, is it correct to understand that this applies to the character count of the input prompt, and the output character count is not relevant and is always limited to around 4K tokens?