Posts say Anthropic makes Fable 5 permanent on Claude Max and Team Premium at 50% limits

• Fable 5 becomes a bundled Claude benefit only for Max and Team Premium on July 20, with usage capped at 50% of plan limits, according to Claude's announcement.
• Pro and Team Standard move to credit-based Fable access and get a one-time $100 credit, the split koltregaskes and rohanpaul_ai both summarized.
• Anthropic framed the move as a capacity problem after several staged extensions, while trq212's follow-up said the goal was avoiding a Fable gap on Max.
• A 30-minute entitlement bug hit first: Fable returned after ClaudeDevs's resolution note said users may need to restart Claude Code and reselect the model with /model.
• Claude Code also got a separate concession, with weekly limits 50% higher through August 19 for Pro, Max, Team, and seat-based Enterprise users, according to ClaudeDevs.

Simon Willison's note called the outcome the end of the “Fablepocalypse,” but also pointed out that the $20 Pro plan still does not include bundled Fable. Anthropic's July 1 redeployment note shows why the July 20 line existed: Fable returned after export controls were lifted with only temporary 50% weekly access for Pro, Max, Team, and select Enterprise plans. Moonshot's archived Kimi K3 tech blog landed two days before the reversal with a 2.8T-parameter open model, a 1M-token context window, and explicit comparisons to Fable 5 and GPT-5.6 Sol.

The new access matrix has two rows:

• Max and Team Premium: Fable 5 included from July 20 at 50% of plan limits.
• Pro and Team Standard: Fable 5 available through usage credits, with a one-time $100 credit.
• Rationale: demand was “challenging to predict”; access had been rolled out in stages while Anthropic secured capacity.
• Follow-up framing: Anthropic wanted “more certainty” and called Fable standard at 50% usage for the plans that use it most intensively, according to Claude's follow-up.

The pricing split maps directly to subscription tiers. The Claude pricing page lists Pro at $20 monthly and Max from $100, with 5x or 20x more usage than Pro.

For engineers treating Fable as a coding-agent substrate, the $100 credit is a bridge to metered use, not a bundled allocation.

The Decoder read the July 20 change as a double cut for Max and Team Premium: the bonus usage phase ends the same day, regular limits drop by 33%, and Fable then applies to 50% of those reduced limits.

Community replies immediately priced the credit against actual usage:

• A $100 credit would not last long under heavy Fable use, one reply argued.
• Fable at 50% on Max was framed as “90%+ cheaper than API rates” by doodlestein.
• Pro users still have Fable available through credits, but omar sar's clarification noted that bundled access is the missing piece.

Before the permanent-access announcement, users hit the exact cutover path early: Fable selected inside Claude Code, then “Usage credits are required for this model,” as shown in bridgemindai's screenshot.

The GitHub issue described the same error while subscription usage remained, plus reports that Fable vanished from usage stats and failed on the web as well as Claude Code. Claude Status later described an erroneous usage-credit requirement across Claude.ai, Claude Code, and other surfaces.

ClaudeDevs later said Fable had been unavailable as a selectable model for 30 minutes. If extra usage was enabled, affected users may have been overcharged and would receive refunds plus matching credit, according to ClaudeDevs's resolution note.

Anthropic's public explanation was capacity. The more operational version was messier: trq212 said many people at Anthropic worked around the clock and that it was not clear the change could be finished in time.

trq212's reply to a Max user added two details:

• The goal was no gap in Fable access on Max plans.
• Demand had been “really high,” and bringing Fable to subscriptions full time had taken a lot of work.

Communication remained the sore spot. giffmana told trq212 the communication had been awful after several perceived rug pulls, and trq212's reply conceded there was “lots for us to improve.”

Anthropic cited capacity, but the reaction kept pointing at competitors. Competition from GPT-5.6 Sol and Kimi made API-only Fable untenable, Simon Willison's note argued.

Theo's anonymized T3 Code analytics said Claude overtook Codex when Fable returned to the Claude Code subscription plan, then Codex pulled ahead again after GPT-5.6. The chart in Theo's post put Codex at 58.6% of turns and Claude at 41.4%.

Kimi was the cheaper pressure point. bridgemindai said they cancelled two $200 Claude Max subscriptions and bought one $199 Moonshot Vivace plan after Kimi K3 did better at frontend design.

The capability pressure was not only vibes. In doodlestein's FrankenGraphDB test, Kimi K3 reviewed a 1.45 MB plan, then Fable spent 44 minutes and roughly 2.3M verification tokens adjudicating that review with a 19-agent workflow.

Tooling caught up fast: an OpenCode post said Kimi K3 could be selected by authenticating a Moonshot API key and choosing Kimi K3 from /models.

Fable was not the only concession. ClaudeDevs said weekly Claude Code limits stay 50% higher through August 19 for all Pro, Max, Team, and seat-based Enterprise users.

TestingCatalog framed it as “50% more testing” after Fable stayed on Max and Team Premium in its Claude Code post.

The access layer touched more than chat. It surfaced inside Claude Code sessions in bridgemindai's terminal screenshot, compaction in daniel_mac8's report, and model selection in ClaudeDevs's resolution note.