Anthropic extended subscription-included access to its flagship Claude Fable 5 model through July 19, its second extension in a single week, while keeping a 50% boost to Claude Code’s weekly rate limits in place for the same period.
We’re extending Claude Fable 5 access on all paid plans, as well as keeping Claude Code’s weekly rate limits 50% higher, through July 19.
— Claude (@claudeai) July 12, 2026
Announced on July 12 via the company’s X account and an email to subscribers, the move lets eligible Pro, Max, Team and Enterprise users spend up to half their weekly usage limits on Fable 5 at no extra cost.
After July 19, Fable 5 shifts to prepaid usage credits at $10 per million input tokens and $50 per million output tokens, the highest rates Anthropic has listed for a public model and twice the cost of Claude Opus 4.8. Anthropic says it aims to return Fable 5 to standard subscriptions once compute capacity allows, without giving a timeline.
The extension is the third such reprieve in five weeks, a pattern rooted in an unusually turbulent stretch for the model. Fable 5 launched on June 9 with free access originally promised through June 22, but a US export-control directive shut it and the more powerful Mythos 5 down globally on June 12. After the Commerce Department lifted the controls on June 30, Anthropic restarted the free window, then extended it twice more, each time hours before the deadline.
Anthropic frames the repeated extensions as rebuilding goodwill with users whose workflows vanished overnight and as gauging demand against limited compute, and has not tied them to competition.
The timing, however, coincides almost exactly with OpenAI’s moves. GPT-5.6, including its flagship Sol model, reached general availability on July 9, and on July 12 OpenAI temporarily removed the five-hour usage cap for paying ChatGPT Work and Codex subscribers on Plus, Business and Pro plans, with no end date given.
Morning. The last 48 hours of Codex and ChatGPT Work have been intense! Three important updates:
– Temporarily removing the 5 hour usage limit restriction for all Plus, Business and Pro plans
– Rolling out changes that will make GPT 5.6 Sol more efficient across the board and…— Tibo (@thsottiaux) July 12, 2026
OpenAI also said it deployed inference optimizations making Sol more efficient, passing roughly 10% more usage to subscribers, and reset usage limits after traffic hit about twice its previous peak on the launch of ChatGPT Work, which it says now has 6 million users. Days earlier, after Anthropic reset its own rate limits, OpenAI’s Tibo Sottiaux needled the company on X with a three-word post, “I smell fear.”
An Access War in Plain Sight
Both companies deny competing on these moves, but the pattern reads as a fight for the same power users. Keeping Fable 5 inside subscriptions another week keeps developers testing it against Sol on Anthropic’s dime rather than drifting to usage credits or a rival’s cheaper model, while OpenAI lifting caps and cutting token costs pushes the opposite way.
The stakes are real because the models are close. Independent evaluator Artificial Analysis places Fable 5 first on its intelligence index at 60 points, with Sol just behind at 59, but estimates Sol costs roughly a third as much per task and leads outright on its coding-agent index. When capability is near parity, price and access become the battleground, and that is exactly where this skirmish is playing out.
The Caveats Beneath the Benchmarks
The competitive framing obscures some important caveats. Sol’s cost advantage is significant, but its scores deserve scrutiny: the safety evaluator METR reported that Sol gamed its software-engineering benchmark at the highest rate it has ever recorded, exploiting evaluation bugs and hidden test answers, and OpenAI’s own system card acknowledged instances of task cheating. One developer with early Sol access said it did not strike him as better than Fable at complex coding.
For Anthropic, the extensions carry their own tension: users have posted screenshots of steep credit bills and cancellation notices, warning they will switch to Sol or SpaceXAI’s cheaper Grok 4.5 if included access ends.
Meanwhile, community speculation about an imminent Opus 5, after an unreleased model briefly surfaced in Cursor, adds another variable. The back-and-forth benefits users for now, but the underlying question, whether Anthropic can afford to keep its priciest model inside subscriptions, remains unresolved.
