Anthropic has pushed back the shutdown date for Fable access on its paid Claude plans again, and according to Simon Willison, the reason is simple: OpenAI just shipped a model that forced its hand. Willison reports that GPT-5.6 Sol is “clearly a Fable/Mythos class model,” and that competitive pressure has Anthropic extending Claude Fable 5 access on all paid tiers through July 19. This is the latest in a string of last-minute reprieves for a model users were told repeatedly would be going away.
What Actually Changed
Anthropic’s new terms, as quoted by Willison, keep Fable 5 alive a little longer and soften the limits along the way:
- Fable 5 access stays on all paid plans through July 19.
- Claude Code’s weekly rate limits remain 50% higher.
- You can spend up to half your weekly usage limit on Fable 5.
- After that, you either keep going with usage credits or switch to another model to stay inside your remaining limits.
The original logic for restricting Fable was compute. Anthropic wanted a clearer read on demand and available capacity before committing to keep a powerful new model cheap for subscribers. That’s a reasonable position for any lab watching its GPU budget. The problem is that every extension makes the policy look less like a plan and more like a flinch.
Why OpenAI’s Move Matters
The contrast with OpenAI is the real story here. Willison points to a morning update from OpenAI’s Thibault Sottiaux, who described “intense” activity across Codex and ChatGPT Work over the prior 48 hours. OpenAI’s response to demand went the opposite direction from Anthropic’s:
- Temporarily removing the 5-hour usage limit for all Plus, Business, and Pro plans.
- Rolling out efficiency changes to GPT-5.6 Sol so it burns less usage and “can take you further,” with exact impact still to be quantified.
- Hitting 6 million active users and landing a usage reset within the hour.
OpenAI, in other words, is loosening limits and projecting confidence while Anthropic keeps its users guessing. When you’re deciding where to build, that difference in posture carries real weight. Nobody wants to wire a workflow around a model that might vanish next Friday.
The Bigger Picture
What stands out is how much of this is about certainty, not raw capability. Both models are in the same performance class, by Willison’s read. The tiebreaker isn’t benchmarks. It’s whether a developer trusts that the tool they picked today will still be there next month. Willison’s take is blunt: he thinks Anthropic should change track and keep Fable permanently available on those plans, because “OpenAI are winning users simply due to the uncertainty that surrounds Fable access.”
That’s a useful reminder for anyone building on top of frontier models. Access policy is a product feature. A slightly weaker model with a stable, predictable plan can beat a stronger one that comes with an asterisk and a countdown clock. Labs compete on tokens per second and reasoning quality, sure. They also compete on whether you can plan around them.
What to Watch
July 19 is the current cliff, but the pattern suggests it may move again. A few things worth tracking:
- Whether Anthropic converts these repeated extensions into a permanent Fable tier, which would end the churn.
- How OpenAI’s efficiency gains on GPT-5.6 Sol actually land once the numbers are published.
- Whether the 6M active user figure holds, since usage resets and lifted limits tend to inflate short-term spikes.
For practitioners, the practical advice is unchanged: if you’re leaning on Fable 5 for anything time-sensitive, keep a fallback model wired up and don’t assume the July 19 date is final. The deadline has slipped before.
More detail, including the full quotes from both companies, is available at the original source.