Update, November 1, 2026: Anthropic’s own Claude Mythos page carried the name Mythos 5.5. In the Internet Archive’s copy from October 31 at 21:49 UTC, the page’s hidden header and logo alt text read “Claude Claude Mythos 5.5” while everything a visitor could see said 5.1; by 20:05 UTC on November 1 the field read Claude Mythos 5.1 again. Anthropic says Mythos 5.1 and Fable 5.1 are one model with different safeguards, so this is the first sign on an Anthropic page that the next pair carries 5.5. Separately, the Azure entry from October 13 was removed from LiteLLM’s price file on October 14. Anthropic’s release notes and OpenRouter still list no 5.5 Fable or Mythos. The record is in the Mythos page section below.
Update, October 28, 2026: An unannounced model named Obsidian 3 appeared in the public data behind Cognition’s FrontierCode 2.0 leaderboard on its launch day, run in the claude-code harness, and scored 46.8% at high effort against 45.5% for Claude Opus 5.5 at max. It was gone from the file by 13:21 UTC on October 28. X posts read it as Fable 5.5; the file itself gives Obsidian 3 the same chart colour as Meta’s Muse Wonder 1.0, not the orange of every Claude model. The record is in the Obsidian 3 section below.
Update, October 19, 2026: The Information reported at 14:00 UTC that Anthropic is readying a new model, under the headline “Anthropic Readies Model That Reads Users’ Intentions Better”. The story is paywalled. TestingCatalog’s summary says the model is due this month and quotes the report: “It’s not clear which model family Anthropic will fall into, such as Fable or Opus.” So the report is about a model, not a name. As of 19:11 UTC Anthropic’s docs, pricing, release notes and OpenRouter still list no Fable 5.5.
Update, October 13, 2026: Fable 5.5 has surfaced in a cloud catalog for the first time. LiteLLM’s price-sync bot added azure_ai/claude-fable-5-5 at 12:27 UTC, sourced from Microsoft Azure’s East US 2 model list, and an X post carried it an hour later. 30 of the entry’s 31 fields match the Fable 5.1 entry, prices included; only the source line differs. As of 15:53 UTC Anthropic’s docs, pricing and release notes still list no Fable 5.5.
Fable 5.5 is a rumor with a better number than it had two weeks ago. On September 22, 2026 Anthropic released Claude Opus 5.5 and named what comes next: “Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks”. Fable is not in that sentence. As of September 23, 2026, 14:40 UTC, Anthropic’s model documentation lists claude-fable-5-1 as the current Fable, and no Fable 5.5 identifier, price or system card exists.
What does exist is a story that has now worn two numbers. The pretraining-run rumor that circulated as Fable 5.2 in early September was posted as Fable 5.5 two days later, and this week Anthropic moved its whole family to 5.5. Our Fable 5.2 tracker keeps the record of the 5.2 claims; this page tracks the 5.5 label from here.
On this page · 14 sectionsOpen
- What we actually know
- The same story, two numbers
- The Mythos page, October 31
- The Azure catalog entry, October 13
- The Information report, October 19
- Obsidian 3 in Cognition’s FrontierCode data, October 27
- The bar Opus 5.5 just set
- What Anthropic’s own rhythm says
- What would confirm it
- Where CellCog stands
- What we are watching for
- The tracker
- Update log
- Sources
On October 19, 2026 The Information reported that Anthropic will release a new model this month, one especially good at reading a person’s intentions. The report did not say whether it is a Fable or an Opus. Anthropic’s docs still list no Fable 5.5; an Azure catalog entry from October 13 was pulled a day later, and on October 31 Anthropic’s own Mythos page carried a hidden Mythos 5.5 label.
The story has worn two numbers. On September 8 @kimmonismus relayed a new pretraining run as the next Fable, calling the name provisional; our Fable 5.2 tracker holds that chain. On September 10 @Mr_Salio posted the same run as Fable 5.5, late September or early October.
The 5.5 label is now the likelier one, by inference only. Anthropic gave Opus, Sonnet and Haiku the number 5.5 this week, so a next Fable carrying 5.5 would match the family. Anthropic has not said a Fable is coming at all.
The bar moved on September 22. Anthropic says Opus 5.5 performs at the level of Fable 5.1 on most work at $4 and $20 per million tokens, against Fable 5.1’s $10 and $50. A Fable 5.5 has to justify the premium over a model that already matches its predecessor.
Fable 5 to Fable 5.1 took 84 days. A Fable 5.5 shipping by October 31 would come 60 days after 5.1. CellCog moved every Claude tier from Fable 5.1 to Opus 5.5 on September 22; a Fable 5.5 would be a same-day routing decision, and this page becomes its record.
§ 01What we actually know
| Claim | Status |
|---|---|
| A new Fable pretraining run exists | X posts, September 8 and 10; no Anthropic document |
| It will be called Fable 5.5 | @Mr_Salio, August 13 and September 10; the September 8 relay called the name provisional |
| Late September or early October release | @Mr_Salio, September 10; no release by October 19 |
| Most capable Anthropic model yet | @Mr_Salio, August 13; no benchmark |
| Keeps Fable 5 pricing | @Mr_Salio, August 13; no price sheet |
| Deployment slowed by restrictions | @Mr_Salio, August 13; no Anthropic statement |
| Announced with Opus 5.5 | No; Anthropic named Sonnet 5.5 and Haiku 5.5 only |
| A new Anthropic model this month | The Information, October 19, as relayed by TestingCatalog; family not named |
| Anthropic identifier, system card, price | None exist |
| Listed in a cloud catalog | Azure East US 2 entry added October 13 with Fable 5.1’s specs; removed from LiteLLM October 14 |
| Named on an Anthropic page | Mythos 5.5 in the Mythos page’s hidden header, archived October 31; reads 5.1 again November 1 |
§ 02The same story, two numbers
| Date | Event | Source |
|---|---|---|
| Aug 13, 13:44 | Fable 5.5 expected to become Anthropic’s most capable model | @Mr_Salio on X |
| Sep 1 | Claude Fable 5.1 released | Anthropic |
| Sep 5 | Fable 5.2, or whatever it ends up being called, expected very soon | @kimmonismus on X |
| Sep 8 | New pretraining run, end of September or early October | @kimmonismus on X |
| Sep 10, 11:58 | Fable 5.5 in late September or early October, first fresh pretraining run | @Mr_Salio on X |
| Sep 18 | Fable 5.1 reported routing to a newer Fable in Claude Code | @notjazii, @chetaslua on X |
| Sep 22, 16:31 | Opus 5.5 released; Sonnet 5.5 and Haiku 5.5 named as next | Anthropic |
| Sep 23, 14:40 | Docs, release notes, pricing, OpenRouter: no Fable 5.5 or 5.2 | Our read |
| Oct 13, 12:27 | azure_ai/claude-fable-5-5 added from Azure East US 2 | LiteLLM price-sync bot |
| Oct 13, 13:33 | Catalog entry posted, 74K views by 15:52 UTC | @alexgetmancom on X |
| Oct 13, 15:53 | Docs, release notes, pricing, OpenRouter: still no Fable 5.5 | Our read |
| Oct 14, 16:17 | Entry removed: Azure does not list it | LiteLLM commit |
| Oct 19, 14:00 | New Anthropic model reported due this month, family unnamed | The Information |
| Oct 19, 14:52 | Report relayed with quotes; a second post cites Fable 5.5 traces | @testingcatalog on X |
| Oct 19, 19:11 | Docs, release notes, pricing, OpenRouter: still no Fable 5.5 | Our read |
| Oct 27, 19:48 | Obsidian 3 in FrontierCode 2.0 data, claude-code harness, top score | Cognition data file, archived |
| Oct 28, 13:21 | Obsidian 3 removed from the data file | Internet Archive copy |
| Oct 31, 21:49 | Mythos page hidden header reads Claude Claude Mythos 5.5 | Internet Archive copy |
| Nov 1, 15:23 | View-source screenshot of the label posted | @kimmonismus on X |
| Nov 1, 20:05 | Page reads Claude Mythos 5.1 again; no 5.5 in release notes or OpenRouter | Our read |
On September 5 @kimmonismus wrote “Fable 5.2, or whatever it ends up being called, is expected very soon”, and on September 8 relayed a new pretraining run with an “End of september or early october release planned”. Two days later @Mr_Salio posted: “Fable 5.5 could launch in late September or early October” and “First fresh pretraining run for the Fable class”. Same run, same window, a different number. The earlier August 13 post had used 5.5 already, adding “Deployment has reportedly been slowed by restrictions”.
We saw the same pattern with Opus: the Opus 5.2 rumor was renamed Opus 5.5 on September 20, and two days later Anthropic shipped a model called Opus 5.5. That is one precedent, not a law. The honest reading today: if Anthropic ships a new Fable soon, 5.5 is the more likely label because the rest of the family now carries it. Whether a new Fable is coming soon at all has no Anthropic evidence.
§ 03The Mythos page, October 31
Anthropic’s Claude Mythos page is the product page for its restricted model class, the one vetted cyberdefenders and life scientists can use. Its headline is a logo image, and the page carries the model name in two places a reader never sees: a hidden header for screen readers and the image’s alt text. In the Internet Archive’s copy from October 31, 2026 at 21:49 UTC, both read “Claude Claude Mythos 5.5”, six times across the HTML and its data payload. Everything a visitor could see still said 5.1: the logo image, the tagline and the newest announcement on the page, Introducing Claude Mythos 5.1 from September 1.
| Copy | Hidden header and alt text | Logo image file |
|---|---|---|
| Archive, Oct 23, 02:24 UTC | Claude Claude Mythos 5 | Unchanged |
| Archive, Oct 31, 21:49 UTC | Claude Claude Mythos 5.5 | Unchanged |
| Our read, Nov 1, 20:05 UTC | Claude Mythos 5.1 | Unchanged |
On November 1 at 15:23 UTC @kimmonismus posted a view-source screenshot of the label, crediting an anonymous tip, and the post passed 270,000 views in under five hours. @AiBattle_ then showed that the October 23 copy did not have it. When we read the page at 20:05 UTC, from three places (a plain request, our own browser and a second browser), it said Claude Mythos 5.1 in all of them, and the doubled word was gone too. We saved a fresh archive copy at that minute.
Two readings fit. The site may have been prepared for a Mythos 5.5 launch and the field pulled once people noticed. Or an editor typed the wrong version into a field only screen readers use, then fixed it along with the extra Claude. The October 23 copy said Mythos 5, not 5.1, which shows the field had been out of date before. Neither reading comes with a model, a price or a date.
It matters for Fable because the two names share one model. Anthropic’s launch page for the current pair says “Claude Fable 5.1 and Claude Mythos 5.1 are the same model, but with different levels of safeguards.” If Anthropic pairs them again, a Mythos 5.5 means a Fable 5.5. That makes 5.5 the likeliest number for the next pair, and it is still not an announcement.
§ 04The Azure catalog entry, October 13
On October 13, 2026 at 12:09 UTC LiteLLM’s provider-sync bot opened pull request #46201, “feat(azure): add azure_ai/claude-fable-5-5”, and it merged at 12:27 UTC. LiteLLM, an open source model gateway, keeps a public price file, and the bot fills it from what each cloud reports. The entry’s source line points at Azure’s management API listing models for the East US 2 region, a list only an Azure subscriber can read. The next day, October 14 at 16:17 UTC, a registry audit removed it with the commit message “remove azure_ai/claude-fable-5-5, which Azure does not list”.
| Field | Fable 5.5 entry | Fable 5.1 entry |
|---|---|---|
| Input, per million tokens | $10 | $10 |
| Output, per million tokens | $50 | $50 |
| Cache read, per million tokens | $0.25 | $0.25 |
| Context window | 1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Deprecation date | December 5, 2027 | December 5, 2027 |
Of the 31 fields in the entry, 30 match the Fable 5.1 entry; the only difference is the source line. Two readings fit. Microsoft has a Fable 5.5 deployment staged in its catalog ahead of an announcement, or a catalog row was created early with its predecessor’s values as placeholders. Either way the figures say nothing about what Fable 5.5 costs or does.
The X post that spread the entry also reported a GenerallyAvailable status in Azure’s list. LiteLLM’s entry records no status field, and we cannot read the Azure list itself, so that part is the poster’s account. A reply from @LuminaBench noted “the pricing/context/status and stuff are copied from fable 5.1”; the poster agreed and added that “the model itself has actually shown up”. What the entry does settle is the name: the 5.5 label @Mr_Salio has posted since August 13 is the one in Microsoft’s catalog.
§ 05The Information report, October 19
The Information published “Anthropic Readies Model That Reads Users’ Intentions Better” by Stephanie Palazzolo at 14:00 UTC on October 19, 2026. The article is paywalled; we read its headline and timestamp, not its body. The lines below come from TestingCatalog’s summary on X, which quotes it.
| Point | As relayed |
|---|---|
| Timing | Release this month |
| Main strength | Inferring a person’s intentions and goals from little information |
| Second strength | Thinking outside the box instead of common-sense advice |
| Model family | Not clear, Fable or Opus |
| Model name | None given |
| Anthropic statement | None reported |
TestingCatalog quotes the report: “The unreleased model is especially good at discerning a person’s underlying intentions and goals even if they don’t give the model much information.” And: “The Anthropic model is also better at thinking outside of the box rather than regurgitating common sense advice or solutions.” The same post places the release alongside Gemini 4 Carbon from Google and a new Terra model from OpenAI.
Two lines in that thread matter for this page. The report: “It’s not clear which model family Anthropic will fall into, such as Fable or Opus.” And TestingCatalog’s own follow-up a few seconds later: “many Fable 5.5 traces have been observed recently.” The first says The Information did not name a Fable. The second is TestingCatalog’s reading, which points back to the Azure entry and the routing reports. Write-ups that headline this as Fable 5.5 are joining the two.
§ 06Obsidian 3 in Cognition’s FrontierCode data, October 27
Cognition launched FrontierCode 2.0, its coding benchmark graded on whether a maintainer would merge the change, on October 27, 2026 (our read of the index it feeds). Its leaderboard reads from a public data file. The Internet Archive’s copy from 19:48 UTC that day lists 26 configurations, one of them Obsidian 3, a name no lab has announced. An X post at 21:10 UTC spotted it. In the next archived copy, 13:21 UTC on October 28, the entry is gone, and the live file now lists 24 models.
| Effort | Obsidian 3 score | Obsidian 3 cost per task | Claude Opus 5.5 score | Claude Opus 5.5 cost per task |
|---|---|---|---|---|
| low | 39.2% | $2.20 | 14.7% | $0.75 |
| medium | 39.8% | $2.84 | 30.1% | $1.70 |
| high | 46.8% | $8.62 | 34.5% | $2.42 |
| xhigh | 45.9% | $19.59 | 43.4% | $5.18 |
| max | 45.5% | $47.44 | 45.5% | $12.82 |
The scores are the file’s own figures; Cognition has published nothing about the entry. Two clues point in opposite directions.
- Toward Anthropic: the harness. Every other claude-code entry in the file is a Claude model (6 of them), while OpenAI’s models run in Codex, Grok 4.7 in Grok Build and Muse Wonder 1.0 in Meta’s own harness.
- Away from Anthropic: the colour. The file assigns each model a chart colour; all 6 Claude entries share Anthropic’s orange, and Obsidian 3 carries the same blue as Muse Wonder 1.0.
It is also expensive. At high effort Obsidian 3 cost $8.62 a task, 3.6 times Opus 5.5’s $2.42, and its max-effort run cost $47.44. As of 19:01 UTC on October 28, Anthropic’s release notes show nothing newer than October 23, and neither Anthropic nor Cognition has commented.
§ 07The bar Opus 5.5 just set
| Model | Input | Output | Anthropic’s description |
|---|---|---|---|
| Claude Fable 5.1 | $10 | $50 | current Fable, September 1 |
| Claude Opus 5.5 | $4 | $20 | at the level of Fable 5.1 on most work |
Anthropic’s own sentence is the problem a Fable 5.5 has to solve: Opus 5.5 “performs at the level of Claude Fable 5.1 on most work”. A premium Fable has to clear a model that already matches the last premium Fable at 40 percent of its output price. The August 13 post said Fable 5.5 would keep Fable 5 pricing; if so, the gap it has to earn is 2.5 times on output.
§ 08What Anthropic’s own rhythm says
Fable 5 shipped June 9, 2026 and Fable 5.1 on September 1, 84 days apart. A Fable landing by October 31, inside the month The Information reported, would come 60 days after Fable 5.1, faster than the 84-day turn before it. Across Anthropic’s whole lineup the pace is faster, and Opus 5.5 came 21 days after Fable 5.1, so a model in October is within the company’s rhythm; a Fable specifically in October would be a first. October ended without one.
§ 09What would confirm it
- An Anthropic announcement page or news entry for a new Fable.
- A new claude-fable identifier in the model documentation.
- A system card at anthropic.com.
- A catalog line on Bedrock, Google Cloud, Foundry or OpenRouter. Azure’s East US 2 list was read as carrying one on October 13, with specs copied from Fable 5.1, and the entry was pulled a day later; a line with its own prices would count.
- A Fable 5.5 or Mythos 5.5 name in visible text on an Anthropic page. A hidden label that was changed back does not count on its own.
- A dated statement from Anthropic naming the model.
A benchmark screenshot, a Claude Code routing report or another post with a window would not. Those go in the chain table.
§ 10Where CellCog stands
CellCog ran its Core and Max tiers on Claude Fable 5.1 from its release day and moved every Claude tier to Claude Opus 5.5 on September 22 (what changed), on price and on behavior in long agentic runs. A Fable 5.5 would be a same-day routing decision on the same terms, and the decision would land here.
§ 11What we are watching for
- Anthropic’s news, model docs, pricing and release notes on every scan.
- Provider catalogs for a new claude-fable identifier.
- A second source with an artifact: a string, a model-list screenshot, a system card.
- Whether the 5.2 label resurfaces or the 5.5 label holds.
- Whether Mythos 5.5 returns to Anthropic’s Mythos page, this time where readers can see it.
- A Bedrock, Vertex or Foundry identifier for a Fable or Mythos 5.5.
- Whether the model The Information described ships as a Fable, an Opus or something else, now that October has ended without it.
- Whether Obsidian 3 returns to Cognition’s leaderboard under a public name.
§ 12The tracker
As of November 1, 2026, 20:05 UTC: no Fable 5.5 or Mythos 5.5 from Anthropic. Its Mythos page showed a hidden Mythos 5.5 label in an October 31 archive copy and reads 5.1 again; the Azure catalog entry from October 13 was removed a day later. No new Fable or Mythos shipped in October.
§ 13Update log
As of November 1, 2026, 20:05 UTC: Anthropic’s Mythos page carried Claude Claude Mythos 5.5 in its hidden header and alt text in an October 31 archive copy; the live page reads Claude Mythos 5.1 again. LiteLLM’s Azure entry was removed October 14. Section, chain rows and tracker updated. Retitled.
As of October 28, 2026, 19:10 UTC: an unannounced model, Obsidian 3, sat in Cognition’s FrontierCode 2.0 data on October 27 in the claude-code harness with the top score, then was removed by October 28; its chart colour matches Meta’s Muse Wonder 1.0. Section and chain rows added. Retitled.
As of October 19, 2026, 19:11 UTC: The Information reported a new Anthropic model due this month (14:00 UTC); TestingCatalog’s relay quotes it as unclear whether the model is a Fable or an Opus. Docs, pricing, release notes and OpenRouter list no Fable 5.5. Retitled.
As of October 13, 2026, 15:53 UTC: azure_ai/claude-fable-5-5 appeared in LiteLLM’s price file at 12:27 UTC, synced from Azure East US 2, every spec copied from Fable 5.1. Anthropic docs, pricing, release notes and OpenRouter still list no Fable 5.5. Retitled.
As of September 23, 2026, 14:40 UTC: page opened. Rumored window late September or early October.
§ 14Sources
- Anthropic, Claude Mythos page, archived October 23, 02:24 UTC, October 31, 21:49 UTC and November 1, 20:05 UTC; live page read November 1, 2026, 20:03 to 20:06 UTC.
- Anthropic, Claude Fable 5.1 and Mythos 5.1, September 1, 2026.
- LiteLLM, pull request #46428, October 14, 2026.
- X: @kimmonismus, November 1 and @AiBattle_, November 1. Read directly; times from the post ids.
- Cognition, FrontierCode 2.0 data file, archived October 27, 19:48 UTC and October 28, 13:21 UTC; live file read October 28, 2026, 19:05 UTC.
- Anthropic, Introducing Claude Opus 5.5, September 22, 2026.
- Anthropic, models overview, pricing and release notes, read October 19, 2026, 19:11 UTC.
- OpenRouter, Anthropic catalog, read October 19, 2026, 19:11 UTC.
- LiteLLM, commit 7dd53d6 and pull request #46201, October 13, 2026.
- The Information, Anthropic Readies Model That Reads Users’ Intentions Better, Stephanie Palazzolo, October 19, 2026, 14:00 UTC. Paywalled; headline and time read directly.
- X: @testingcatalog, October 19 and its follow-up. Read directly; times from the post ids.
- X: @alexgetmancom, October 13 and its replies. Read directly; times from the post ids.
- X: @Mr_Salio, August 13 and September 10; @kimmonismus, September 5 and September 8. Read directly; times from the post ids.
Q1Where does the Fable 5.5 rumor come from?
X posts from @Mr_Salio. On August 13, 2026 the account said Fable 5.5 was expected to become Anthropic’s most capable model and that deployment had been slowed by restrictions. On September 10 it said Fable 5.5 could launch in late September or early October as the first fresh pretraining run for the Fable class. No document or named source accompanies either post.
Q2Is Mythos 5.5 real?
Only as a label so far. Anthropic’s Claude Mythos page showed Mythos 5.5 in hidden text in an October 31, 2026 archive copy and read Mythos 5.1 again by November 1. No Mythos 5.5 identifier, price or system card exists.
Q3What would Fable 5.5 cost?
Unknown. The August 13 post said it would keep Fable 5 pricing, and the October 13 Azure catalog entry lists $10 and $50 per million tokens, but every field in that entry is copied from Fable 5.1. Claude Fable 5.1 lists at $10 input and $50 output; Opus 5.5 lists at $4 and $20.
Q4What did the Azure catalog leak show?
An identifier, azure_ai/claude-fable-5-5, in Microsoft Azure’s East US 2 model list, which LiteLLM’s price-sync bot copied into its public price file on October 13, 2026. Its $10 and $50 prices, 1M context, 128K output and December 5, 2027 deprecation date are the same as Fable 5.1’s entry, so they say nothing about Fable 5.5’s real specs. The name matches the rumor. LiteLLM removed the entry on October 14, 2026, saying Azure does not list it.
Q5Did The Information report Fable 5.5?
No. Its October 19, 2026 report, Anthropic Readies Model That Reads Users’ Intentions Better, describes an unreleased model due this month. TestingCatalog’s summary quotes it as saying it is not clear whether the model is a Fable or an Opus. Relays that call it Fable 5.5 add the name themselves.
Q6Should I wait for Fable 5.5?
Not on this evidence. Opus 5.5 already matches Fable 5.1 on most work at a lower price, and Fable 5.1 is live. For agent work, the layer around the model matters more than the version: memory that survives the conversation, a task board and approvals a human can see.