[BUG] Haiku 4.5 usage being counted the same as Sonnet 4.5 despite 3x lower cost

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Haiku 4.5 requests are consuming my usage limits at the same rate as Sonnet 4.5 requests, negating the cost efficiency benefits of the smaller model. This is particularly frustrating on the Max plan where expanded usage capacity is a core value proposition.

What Should Happen?

When using Haiku 4.5, my usage limits should reflect the model's lower cost. I should be able to make approximately 3x more requests with Haiku 4.5 compared to Sonnet 4.5 before hitting my subscription limits. As a Max plan subscriber paying for expanded usage, this cost efficiency should directly translate to significantly more available interactions.

Error Messages/Logs

Steps to Reproduce

Check initial usage status:

Open Claude Code

Note your current usage limit status (e.g., "X requests remaining this week/month")

Make a request with Sonnet 4.5:

Switch to Claude Sonnet 4.5 model

Make a coding request (e.g., "Write a simple function to calculate fibonacci numbers")

Note how much your usage decreased

Check usage after Sonnet request:

Check your usage limit status again

Record the decrease in available requests

Make a similar request with Haiku 4.5:

Switch to Claude Haiku 4.5 model

Make a similar complexity coding request

Note how much your usage decreased

Compare usage consumption:

Compare the usage decrease between Sonnet 4.5 and Haiku 4.5

Expected: Haiku 4.5 should consume approximately 1/3 the usage of Sonnet 4.5 (since it costs 1/3 as much)

Actual: Both models consume the same amount of usage quota

Alternative Reproduction Method
Monitor usage while exclusively using Haiku 4.5 for a session

Calculate effective token cost based on usage consumed

Compare against documented Haiku 4.5 pricing ($1/$5 per million tokens)

Observe that usage consumption rate matches Sonnet 4.5's pricing ($3/$15 per million tokens) instead

Claude Model

Other

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.0.21 (Claude code)

Platform

Other

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Please update the usage counting logic to properly reflect Haiku 4.5's lower cost, and update the support documentation at https://support.claude.com/en/articles/11145838-using-claude-code-with-different-models to include Haiku 4.5's limits.