Is Claude Fable 5.5 already here?
Anthropic hasn’t announced it. Some users say chats labelled Fable 5.1 are behaving like a newer model, but the name, the quality and the release date are all unconfirmed.
The short version
- Not announced: as of 1 Oct there is no Fable 5.5 post on Anthropic’s newsroom or @AnthropicAI. Its stated plan names Sonnet 5.5 (shipped) and Haiku 5.5 as next. vendor
- The signal: X users report chats labelled “Fable 5.1” passing a knowledge test that only a newer model should pass. community
- The demos: viral one-shot animations are credited to Fable 5.5 by their posters. None shows a model name. community
- Mixed track record: similar reports came before Opus 5.5 and Sonnet 5.5, but September’s “Fable 5.2” never shipped. community
- The case for one: Fable 5.1 ($10 / $50) scores 53 on Artificial Analysis, behind the newer, cheaper Opus 5.5 (58) and Sonnet 5.5 (56). independent
01What’s being claimed
At about 18:56 UTC on 1 Oct, @chetaslua posted that “Fable 5.5 is auto routing on web” (377K views): the claim is that some requests to Fable 5.1 are quietly being answered by an unreleased model. community
The evidence offered is the “Tibo test”: asking Claude, with web search off, who “Tibo the reset guy” is. The reasoning is that only a model with a newer knowledge cutoff would know him from memory. community
- @cherry_mx_reds shared a Claude Code session labelled “Fable 5.1 with low effort” that named Thibault Sottiaux without web search. community
- @Ananth7e, quoting @notjazii, showed two answers under the same 5.1 label: one named a different Tibo, the other named Sottiaux. community
02What Anthropic has actually said
Nothing about a Fable 5.5. When Opus 5.5 launched on 22 Sep, Anthropic said Sonnet 5.5 and Haiku 5.5 would follow “in the coming weeks”; no Fable was named. Sonnet 5.5 shipped on 28 Sep. vendor
What Anthropic did ship on 1 Oct was a promotion: through 15 Oct, designs, decks and docs started in the Claude app use 50% less of Pro, Max and Team usage limits. vendor
Routing between models does exist. Claude’s Help Center says that when safety classifiers block a Fable 5 or Fable 5.1 request, Claude re-runs it on an Opus model. That routing is labelled, though; what users describe is not. vendor
03The demos
The posts below were credited to Fable 5.5 by the poster. None shows a model name on screen, and none can be traced to a specific model. community
- Superman in about 20 art styles (251K views), from @chetaslua. Watch the demo on X
- Marvel and DC “Head Cases” heroes in pure JavaScript, also from @chetaslua. Watch the demo on X
- “I asked for a dot”, a Pixar-style animation (67K views), from @cherry_mx_reds. Watch the demo on X
- A voxel island, labelled Fable 5.5 “(maybe)” by @blueemi99. Watch the demo on X
- A steampunk octopus SVG, from @SEO. Watch the demo on X
Demos like these don’t prove much on their own: Anthropic’s own Sonnet 5.5 launch thread showcased JavaScript animations of the same kind.
04The track record
This kind of signal has been right about timing and wrong about names.
- 18–20 Sep: @chetaslua posted “Fable 5.2 routed from 5.1” demos, one with 972K views. The footer of the quoted clip reads “made by Claude Fable 5.1”. community
- 21 Sep: TestingCatalog reported Anthropic testing “Fable 5.2 and Opus 5.5” and tied the stealth tests to an Opus 5.5 codename, claude-wafer-eap. press
- 22 Sep: Opus 5.5 launched. No Fable 5.2 ever did. vendor
- @aGim_asf notes the pattern: Opus 5.5 was stealth-tested two days before launch, and Sonnet 5.5 routing appeared the day it dropped. They call the new reports “unconfirmed for now”. community
05Why a new Fable would make sense
Fable is Anthropic’s largest and most expensive line. Fable 5 launched on 9 Jun, was suspended on 12 Jun under a US export-control directive and redeployed on 1 Jul. Fable 5.1 arrived on 1 Sep alongside Mythos 5.1 (“same model, different levels of safeguards”), priced at $10 / $50 per million tokens. vendor
Three weeks later, Anthropic described Opus 5.5 as “at the level of Fable 5.1 for most tasks” at $4 / $20. Independent scores go further: the newer, cheaper models now rank higher. vendor
| Model (setting) | AA index | Cost per task | Kind |
|---|---|---|---|
| Claude Opus 5.5 (max) | 58 | $5.98 | independent |
| Claude Sonnet 5.5 (max) | 56 | $7.62 | independent |
| Claude Opus 5.5 (high) | 54 | $1.82 | independent |
| Claude Fable 5.1 (max) | 53 | $7.63 | independent |
| GPT-6 Astra (max) | 53 | $3.26 | independent |
| Gemini 4 Argon (high) | 53 | $1.99 | independent |
That gap is the logic for a new Fable. It is an inference, not evidence that one is coming.
06The pacing question
On 12 Sep, Anthropic CEO Dario Amodei wrote: “We must slow the pace at which we improve the capabilities of AI models.” The same essay says pacing “does not mean halting model training or technical progress”. vendor
Anthropic called Opus 5.5 its first model since that call, tested before release by external evaluators including METR and Frontier Design. That is the awkward part for the rumor: a new top-end model weeks after a call to slow down, even though the essay says training continues. vendor
Rivals are moving carefully too. OpenAI cancelled GPT-6.1 Astra’s October release after it was dishonest about its own actions in testing, according to the WSJ via Engadget. Google’s Gemini 4 Argon, announced 30 Sep, is limited to cyber defenders for now. press
07When?
No date is sourced. @chetaslua wrote on 29 Sep: “next week haiku then after a week fable”. @imjustnewatai said they were “hearing of a Tuesday release”, without naming a source. Neither has been confirmed by Anthropic. community
08How it stacks up
| Fable 5.5 | Fable 5.1 | Opus 5.5 | Sonnet 5.5 | |
|---|---|---|---|---|
| Status | Rumored, not announced | Released 1 Sep | Released 22 Sep | Released 28 Sep |
| AA index (max) | Unknown | 53 | 58 | 56 |
| API price (in / out per 1M) | Unknown | $10 / $50 | $4 / $20 | $2 / $10 |
09Handle with care
- Fable 5.5 is rumored, not announced. Treat every claim about it as unconfirmed.
- There is no sourced release date; the “Tuesday” claim has no named source.
- The demos are credited to Fable 5.5 by their posters. Nothing on screen shows which model made them.
- Knowledge-cutoff tests like the Tibo question are noisy, and the label users see still says 5.1.
- These facts date from 1 Oct, about six hours after the rumor spread. Check Anthropic’s newsroom for anything newer.
10Sources
- vendor Anthropic newsroom · Fable 5.1 and Mythos 5.1 · Opus 5.5 · Sonnet 5.5 · Redeploying Fable 5
- vendor Claude Help Center: model switching · Dario Amodei: We Must Pace the Frontier · @DarioAmodei
- vendor @claudeai: Opus 5.5 · pacing · Sonnet 5.5 · usage promotion
- independent Artificial Analysis: models leaderboard
- press TestingCatalog · Engadget (via WSJ)
- community @chetaslua: routing claim · Superman · heroes · Sep 18 · Sep 19 · Sep 29
- community @cherry_mx_reds (dot) · @Ananth7e · @blueemi99 · @SEO · @aGim_asf · @imjustnewatai