By Today is the deadline. Fable 5 free access expires at 11:59 PM PT on Claude Pro, Max, and Team plans — unless Anthropic extends for the fourth time. There is strong evidence they’ve been delaying for a reason: on July 8, an unannounced model called Claude Honeycomb EAP briefly appeared in Cursor’s production model picker, surfaced with real metadata and a selectable UI entry, and vanished within hours. Anthropic has not said a word about it. The specs that leaked point at Opus 5.
What the Cursor Leak Showed
Developer @chetaslua posted screenshots the same day. The Cursor picker listed the model as “Anthropic research model with per-turn controls and safety fallbacks. Early access preview.” Two specifications stood out: a 1-million-token context window and a reasoning mode labeled “extra high effort.”
The most revealing detail is the safety fallback. Honeycomb routes sensitive prompts down to Claude Opus 4.8 for handling — which means, in Anthropic’s architecture, Honeycomb sits above Opus 4.8 in capability. You don’t route queries to a weaker model unless you’re a stronger one. That’s not speculation; that’s how hierarchical fallback logic works.
The EAP label closes the loop. EAP stands for Early Access Program — the exact same tag Anthropic attached to Fable 5 before it launched on June 9. Fable 5 appeared in Cursor roughly two weeks before public release. If that pattern holds, Honeycomb — almost certainly Opus 5 — lands before the end of July.
What Changes Tonight for Developers
Whether or not Opus 5 ships this week, the Fable 5 billing transition is real. Here’s the split:
- Max and Team Premium subscribers: Fable 5 is now a permanent feature at 50% of standard weekly limits. No extra cost. If you’re on Max ($100/month), you’re in the better position.
- Pro and Team Standard subscribers: You receive a one-time $100 usage credit, after which Fable 5 costs $10 per million input tokens and $50 per million output tokens via API billing. The 50% boost to Claude Code weekly rate limits also ends July 20.
That double hit — losing the free Fable 5 window and the rate limit boost simultaneously — is not an accident. Anthropic is consolidating the model under Max-tier value and making the Pro-to-Max upgrade math obvious.
The Honest Cost Comparison
Paying $10/$50 per million tokens is defensible for the right workloads. It is hard to defend when GPT-5.6 Sol hits 88.8% on SWE-Bench Pro versus Fable 5’s 80.3% at half the price ($5/$30). For single-file code generation at high volume, Sol wins on price and on the benchmark.
Where Fable 5 holds up is in long-horizon agentic work: multi-file refactors, architecture decisions, and sessions where the model needs to stay coherent across many hundreds of thousands of tokens. Developers in those workflows have reported $63 per task on Fable 5 against $16 for the same task on Sol. That gap is substantial. The question is whether the coherence difference justifies it for your specific stack.
For high-volume or bulk tasks, Claude Sonnet 5 at $2/$10 per million — five times cheaper than Fable 5, with strong coding performance — handles the majority of what developers actually need Fable 5 for. Anthropic’s pricing page has the full breakdown for building your own cost model.
What to Do Right Now
If you’re on Max or Team Premium, no action required. Fable 5 is yours within your weekly limits.
If you’re on Pro or Team Standard, the $100 credit buys you time but not much runway at Fable 5 rates. Audit your Claude Code usage now: which tasks actually need Fable 5’s long-horizon coherence? Route everything else to Sonnet 5. You’ll spend that $100 slower, and when Honeycomb drops, you’ll be in a better position to evaluate whether Opus 5 pricing justifies the upgrade to Max.
Watch two places for the Opus 5 announcement: anthropic.com/news and the Cursor model picker. That’s where Fable 5 appeared first, and that’s where Honeycomb will reappear when Anthropic is ready. Three extensions in five weeks is not standard operating procedure. Anthropic has been buying time. The Honeycomb leak suggests the wait is close to over.
