Make Fast Mode free for Max subscribers
Preflight Checklist
- [x] I have searched existing requests and this feature hasn't been requested yet
- [x] This is a single feature request (not multiple features)
Problem Statement
Fast mode in Claude Code currently consumes additional usage quota. However, Fast mode uses the same Claude Opus 4.6 model — it just produces faster output. It does NOT switch to a different or cheaper model.
As a Max plan subscriber already paying a premium, being charged extra for faster output of the same model feels unreasonable. Meanwhile, OpenAI's Codex is available to Plus/Pro subscribers at no extra cost, setting a competitive precedent.
Proposed Solution
Include Fast mode as a free perk for Max plan subscribers, similar to how OpenAI bundles Codex with its subscription plans.
Benefits:
- Improved developer experience for paying customers
- Better competitive positioning against Codex
- Increased adoption and loyalty among Max subscribers
- Fair pricing — same model, just faster output, shouldn't cost extra
Alternative Solutions
- Offer a higher Fast mode quota for Max subscribers (e.g., 2x or 3x the current limit)
- Make Fast mode free during off-peak hours for Max subscribers
- Include a generous daily Fast mode allowance as part of the Max plan
Priority
Medium - Would be very helpful
Feature Category
Performance and speed
Use Case Example
As a Max subscriber using Claude Code daily for software development, I frequently toggle Fast mode for quicker iterations. The current usage-based charging for Fast mode adds friction and cost anxiety, discouraging me from using the tool at full speed. With Codex offering free agent access to OpenAI subscribers, I'm considering switching tools.
Additional Context
OpenAI recently launched Codex as a free tool for Plus/Pro subscribers. This sets a strong market precedent. Claude Code's Fast mode — which uses the exact same Opus 4.6 model, just with faster output — should be a free benefit for Max subscribers to remain competitive and reward loyal paying customers.
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗