Anthropic extends Claude Fable 5 access and keeps Code limits high

Anthropic just blinked in the frontier AI price wars, and paying subscribers are the winners.

Anthropic is extending subscription-included access to its flagship Claude Fable 5 model through July 19 at 11:59 p.m. PT. Along with the extension, the company is keeping weekly rate limits for Claude Code bumped by 50% over the same stretch.

The perk applies to Pro, Max, Team, and Enterprise accounts with premium seats. Eligible users can spend up to half of their weekly usage allowance on Fable 5 at no extra charge. Once you hit that 50% ceiling, you can either spend usage credits to keep running Fable 5 or fall back to standard Claude models within your remaining allowance.

Everything updates automatically across Claude Web, desktop, mobile, Code Workbench, and the terminal tool Claude Code. Legacy enterprise tiers are still left out, and API access continues to bill at normal token rates.

The pressure from GPT-5.6 Sol

This marks the second deadline extension for Fable 5. When the Mythos-class model dropped on July 1, Anthropic initially planned to cut off the 50% bundle allowance on July 7.

After that date, users were supposed to shift entirely to usage credits priced at $10 per million input tokens and $50 per million output tokens. That plan has now slipped twice in two weeks.

The real culprit is OpenAI. The broader rollout of OpenAI's GPT-5.6 family, especially Sol, has quickly closed the performance gap. Benchmark tracker Artificial Analysis shows Sol trailing Fable 5 by just one single point on its Intelligence Index while running at significantly lower inference costs. OpenAI has also been actively touting Sol's wins on developer benchmarks like Terminal-Bench 2.1.

Why it matters

If you are a developer with a Claude subscription, take the gift and run your heaviest engineering backlogs right now. Fable 5 is an elite model, but its regular rate of $50 per million output tokens will sting the moment Anthropic turns on the meter. Pairing free access with a 50% bump on Claude Code weekly limits makes this an ideal week for hard refactors.

For the broader industry, this move proves that frontier model dominance no longer guarantees pricing power. Anthropic built an enviable moat among software engineers over the past year, but that loyalty gets fragile when OpenAI ships a model that matches Fable 5 in real-world benchmarks at a fraction of the cost.

In our view, Anthropic is playing defense here. Giving away half your weekly budget on an expensive flagship model eats server capacity and bleeds margin, but pulling the plug right as developers are test-driving GPT-5.6 Sol would risk mass churn. Anthropic bought itself another week of breathing room, but the fundamental problem remains: high inference costs will make July 20 a very awkward transition.