Posts suggest Anthropic adds Fable access to Max plan

• Fable 5 access was extended through July 12, with the Claude UI in LLMJunky's screenshot saying it was included for up to 50% of the weekly usage limit.
• Max became the center of the scramble: thekitze's post said Anthropic was including Fable in Max, while minchoi's Max 20x screenshot showed a separate Fable quota nearly exhausted.
• ClaudeDevs reset 5-hour and weekly rate limits for all users, according to the reset post.
• Builders spent Fable on judgment-heavy work, with shannholmberg's workflow diagram assigning Fable to review and planning before cheaper models execute.

Anthropic's tiny tooltip carried the useful detail: "Extended through July 12" and "up to 50% of your weekly usage limit," captured in LLMJunky's screenshot and alliekmiller's screenshot. Before the extension, ozansihay's model menu showed Fable marked "Included until July 7," with usage credits expected at $10 per 1M input tokens and $50 per 1M output tokens. The creator posts quickly moved from cheering to salvage workflows: shannholmberg's recorder step had Fable write reusable learnings into a repo, while levelsio's Windows 3.11 demo used it to add desktop file drag-and-drop into a browser OS.

thekitze framed the change as Anthropic including Fable in the Max plan, then followed with a join link in a follow-up post.

The broader wording showed up in minchoi's post: Anthropic was "extending access to Claude Fable 5 on all paid plans through July 12" in minchoi's Max 20x screenshot. The in-product tooltip added the cap: included in the plan for up to 50% of weekly usage, as seen again in alliekmiller's screenshot.

That changed the clock from ozansihay's earlier model menu, where Fable was marked "Included until July 7" and the post described standard usage-credit billing after the included period. levelsio turned the extension into the day's best class-war joke in his "underclass" post, while minchoi's graphic called it five more days.

ClaudeDevs posted one blunt line: 5-hour and weekly rate limits had been reset for all users.

Users had found a plan-upgrade reset before that global reset. LLMJunky's Max 5x screenshot showed a fresh 0% usage panel after moving to the $100 plan in the upgrade post, and LLMJunky's reply said a 20-to-100 plan move worked too.

Another LLMJunky reply claimed a friend had tested 10-to-20. By the time the extension screenshots spread, LLMJunky's later reply still described a personal reset arriving 10 hours later.

Fable had its own weekly bar inside the usage UI. In minchoi's screenshot, Max 20x showed Current session at 12%, All models at 53%, and Fable at 99%.

thekitze's dashboard showed a different burn pattern: his usage screenshot had Current session at 18%, All models at 18%, and Fable at 24%. A separate thekitze warning screenshot showed an in-chat notice at 77% of the Fable 5 limit.

One edge case came from JohnMeta8, who said a Max 20x account using Cowork on Sonnet showed 3% Fable usage despite never selecting Fable. thekitze also posted a code-workflow screenshot with Fable 5 1M High selected next to a "commit and push" prompt.

levelsio used Fable to add WebFS drag-and-drop to a Windows 3.11 browser environment: files dropped from the desktop landed in C:\DOCS, and shared files could be moved into C:\SHARED in his demo. In a reply, levelsio described Fable as "like a guy who never gives up."

thekitze's favorite workflow was not polished prompting. In thekitze's voice-dump post, the pipeline was "long and vague voice dump → usable result," which he said had never been better than with Fable.

alliekmiller said she had started prepending a motivation line that made the goal "for you to have fun" in her Fable prompt post, because Anthropic guidance had pushed users to give Fable the "why." marckohlbrugge's ROI thread split expensive Fable work into three buckets:

• Better time and token allocation: Fable reviewed git commits, WIP todos, Linear issues, and AI "board meetings," then produced monthly, weekly, and daily planning sheets.
• Recurring revenue: it scoped a Pro subscription for a job board and onboarding improvements for another app.
• Recurring costs: it improved an eval suite, tested model and prompt variants, and found token-caching and reasoning-effort savings.

shannholmberg's extension-day diagram put Fable above execution: Fable reviews projects and writes plans, then Opus or GPT 5.5 runs the plan in the workflow diagram. The longer playbook used one filter from shannholmberg's post: can a cheaper model redo this tomorrow?

The five-part version treated Fable as a judgment extractor:

• Workspace standard: Fable rewrote CLAUDE.md, skills, and learnings files with conventions, mistakes, and quality bars in shannholmberg's first step.
• Consultant audit: Fable received business context and returned ranked moves, completion criteria, and cheaper-model instructions in shannholmberg's audit post.
• Second brain run: Fable mined niche research, competitors, and recurring problems into atomized linked notes in shannholmberg's research post.
• /goal runs: Claude Code's /goal command set a finish line, while a smaller model checked the stop condition, according to shannholmberg's goal post.
• Recorder skill: a skill fired after each solved problem and wrote a learnings note for later models in shannholmberg's recorder post.

The same thread ended with shannholmberg's closing post claiming that whatever Fable writes down keeps working after the model is gone.