Day 7 · 2026-11-17
35 launches checked147 claims quoted word for word179 user reports linked10 verified fixes linked40 corrections publishedLive 25 days5 of 5 founding sponsor spots left
AI launches, checked against real usersLast re-stamped 17 Nov 2026

All receipts / No. 030

No. 030 · ModelChecked 13 Nov 2026

Day 7 receipt

Claude Fable 5.2 and Mythos 5.2

Anthropic · launched 9 Nov 2026

Claims on the record

We check on 16 Nov 2026

Bottom lineFor most work, Opus 5.5 is still enough; Fable earns its price on 3D and visual demos.

They promised
Up to 30% faster than Fable 5.1, the top score of any released model, and far fewer safety blocks.

Early read, not a stamp

  • The real cost. About 2.6x Opus 5.5 per task in Artificial Analysis tests; four rounds of a 3D film cost $80 on Fable vs $24 on Opus. RoboNuggets (Jay E)
  • Where it wins. Its 3D worlds come out more cinematic and closer to the real place. RoboNuggets (Jay E)
  • Where it falls short. Motion graphics and landing pages come out like for like with the cheaper Opus 5.5. RoboNuggets (Jay E)
  • Still missing. Neither Fable nor Opus writes natural marketing copy yet. RoboNuggets (Jay E)

The claims we’ll check

  • “They generate outputs up to 30% faster than Claude Fable 5.1 and Claude Mythos 5.1”
  • “At Xhigh effort, Fable 5.2 reaches 49.1%, the highest score of any released model we’ve tested.”
  • “we observed about 55% fewer blocks with Fable 5.2 compared with Fable 5.1—about 60% fewer in cybersecurity and 40% fewer in biology.”
  • “a team of 1,000 Fable 5.2 agents fully formalized the theorem in just 30 hours, with minimal human intervention.”
  • “Fable is included for up to 50% of weekly usage limits for Pro plans and standard seats on the Team plan.”

Quoted word for word from Claude Fable 5.2 and Mythos 5.2. Day 4: most figures are still Anthropic's own. One outside lab ranks it first by a narrow margin, and the first side-by-side hands-on test finds it worth the extra cost only for 3D work.

Full evidence 5 claims · 3 user reports · 7 sources

Anthropic released Claude Fable 5.2 to all users and Claude Mythos 5.2, the same model with fewer cyber and biology safeguards, to vetted organizations. It claims faster output, a lower per-token rate, fewer unnecessary blocks, and the top coding score of any released model it has tested.

Launch post: Claude Fable 5.2 and Mythos 5.2

Why this stamp

Day 4: most figures are still Anthropic's own. One outside lab ranks it first by a narrow margin, and the first side-by-side hands-on test finds it worth the extra cost only for 3D work.

Anthropic's announcement, its X account and six outlets describe the same release: Fable 5.2 available to everyone on paid plans, the API and the major clouds, and Mythos 5.2 limited to vetted cyber and life-science organizations. Almost all numbers are Anthropic's own runs, including the 49.1% coding score, the 30% speed gain and the 55% drop in blocked requests. Outside evidence is thin so far. Vals AI ranks the model first of 36 on its overall index, but by under a point over the next model, and Artificial Analysis puts it first on its Intelligence Index while recording sharply higher token use and cost per task at its top effort setting; its per-token rate is lower, so the bill depends on the effort setting. Anthropic's own guidance still recommends its cheaper Opus 5.5 for most work. Early users split: some report big projects finished in days, others say a single prompt used most of a five-hour allowance or that the model is slow. The temporary plan offer for Pro runs to 16 November, so real limit behaviour is not yet visible. We re-stamp on day 7 (16 November) and day 14. Early hands-on (13 Nov): a RoboNuggets side-by-side test with Opus 5.5 found Fable better only on 3D worlds, at several times the cost per task. Still one tester; we stamp on day 7 (16 Nov).

Launch dayWhat they claimed, quoted word for word · 5

“They generate outputs up to 30% faster than Claude Fable 5.1 and Claude Mythos 5.1”
? UnclearEvidence: R3, S1, S2, S5, S6

CNBC repeats the figure, but we found no outside speed test against the 5.1 models. One independent leaderboard run clocked about 46 minutes per task at its top effort setting, a figure that reflects thinking time as well as generation speed.

“At Xhigh effort, Fable 5.2 reaches 49.1%, the highest score of any released model we’ve tested.”
? UnclearEvidence: R3, S3, S4, S5, S6, S7

The score is Anthropic's own run of a benchmark from a partner. VentureBeat notes the headline result needs a deliberate effort setting, since Medium is the default in the apps, and that the score rises with cost per task. An independent lab ranks the model first overall on its own index by a narrow margin, which supports the direction but not this exact number.

“we observed about 55% fewer blocks with Fable 5.2 compared with Fable 5.1—about 60% fewer in cybersecurity and 40% fewer in biology.”
? UnclearEvidence: S5

This comes from Anthropic's early testing. One outside lab reports a zero refusal rate in its run, but it did not compare against Fable 5.1, so the reduction is unchecked.

“a team of 1,000 Fable 5.2 agents fully formalized the theorem in just 30 hours, with minimal human intervention.”
? UnclearEvidence: S6

A single run per model, reported by Anthropic. Commentators note the result is hard to generalise and that independent experiments with many agents found duplicated work and coordination overhead that needed human fixes.

“Fable is included for up to 50% of weekly usage limits for Pro plans and standard seats on the Team plan.”
? UnclearEvidence: R1, R2, S4, S7

Anthropic's launch offer is temporary for Pro. Early users disagree about how fast it drains limits: one user says a single low-effort prompt used most of a five-hour allowance, while another on a top plan says heavy use since launch took a small share of the weekly limit.

Day 7What real users report · 3

R1Reddit10 Nov 2026problem

A user on a Claude Code forum says a single prompt on Fable 5.2 at low effort used about 80% of their five-hour usage, with a context close to a million tokens. Replies point to the large context and to Fable's known appetite.

Thread posted 10 Nov, replies from users only · Open the source

R2Reddit10 Nov 2026praise

A reply in the same thread from a user on the top consumer plan says they have run it at extra-high effort since launch and used only about 10% of their weekly limit, so the drain depends on plan and workload.

Reply in a 10 Nov thread · Open the source

R3Hacker News9 Nov 2026mixed

In the Hacker News thread on the Artificial Analysis result, a top commenter says to ignore the scores and look at cost per task, which they say roughly doubled versus 5.1 at the top setting. Others reply that lower effort settings beat the older models for less.

Thread of 255 comments as mirrored on 10 Nov · Open the source

TestersWhat the testers found · 4

Brock MesarichAI for Non Techies (YouTube)11 Nov 2026

Same prompts on Fable 5.2 and Opus 5.5 across three builds (website, 3D game, video edit): Fable won the game but left a repeated word in the video edit, and its total bill was only about 14% higher despite double list pricing.

“Fable 5.2 costs double on paper, but across three of these tests, it only came out to about 14% more than Opus.” · Open the test

Matthew BermanYouTube9 Nov 2026

After early-access builds (an Excel clone that ran almost 50 hours, a Gears of War-style level), he rates Fable 5.2 above Opus 5.5 but expects to keep Opus 5.5 for everyday work because it is faster and cheaper to run.

“Without a doubt, Fable 5.2 is better than Opus 5.5. However, Opus 5.5 is very efficient.” · Open the test

Alex FinnYouTube9 Nov 2026

Uses Fable 5.2 only to brainstorm, plan and write a task board, then switches to Opus 5.5 at medium effort to build, to save usage.

“If you have really good intelligence, plan out the app, build out the tasks, then you don't need the best intelligence in the world to actually build the app out.” · Open the test

Riley BrownX9 Nov 2026

One prompt, live on stream: Fable 5.2 wrote the Blender code for a 13-shot animated film, rented cloud GPUs and rendered it in 4K.

“It rented 10 GPUs on RunPod and rendered the whole thing in 4K in 52 minutes. Total cost: $5.75.” · Open the test

OthersWhat outlets and testers say · 7

S1The Verge9 Nov 2026neutral

Relays Anthropic's claims of better coding, reasoning and vision and faster output, and notes Fable 5.2 launches today while Mythos 5.2 starts with vetted organizations only.

“Anthropic launches new Claude Fable and Mythos 5.2 models” · Open the source

S2CNBC9 Nov 2026neutral

Confirms the faster-output claim as Anthropic's and quotes a product lead saying Fable is for power use cases, with Opus 5.5 better suited to daily tasks.

“Anthropic releases first frontier models since Amodei's call for slowdown as IPO buzz builds” · Open the source

S3VentureBeat9 Nov 2026positive

Reports strong benchmark leads but stresses the top coding score needs a deliberate higher effort setting and that cost per task can rise even with the lower list rate, so teams should test on their own work.

“Anthropic releases Claude Fable 5.2 and Mythos 5.2, pairing new, record-high benchmarks with 20% API cost reduction” · Open the source

S4Decrypt9 Nov 2026mixed

Describes an AI YouTuber's impressive builds that ran up estimated bills in the thousands, notes the claims need independent scrutiny, and explains the temporary Pro offer ends 16 November.

“Anthropic's Claude Fable 5.2 and Mythos 5.0 Are Impressive, But Be Prepared to Sell a Kidney to Afford Them” · Open the source

S5DEV Community (Hamza Ahmad Aslam)9 Nov 2026mixed

Separates Anthropic's numbers from the one outside lab so far, which ranks it first overall by a narrow margin but mid-table on legal and tax tasks, and says the speed and fewer-blocks claims have no outside check yet.

“Claude Fable 5.2: benchmarks, price and the first independent test results” · Open the source

S6Edward Kiledjian10 Nov 2026skeptical

Says independent benchmarks support the gains but show much higher cost and time per task at maximum effort, and that multi-agent demos do not prove production readiness.

“Claude Fable 5.2 and Mythos 5.2: Progress, Performance and Practical Limits” · Open the source

S7RoboNuggets (Jay E)13 Nov 2026hands-on

Tested Fable 5.2 against Opus 5.5 on real work: 2x the per-token price and about 2.6x per task in Artificial Analysis tests; four rounds of a 3D film cost $80 on Fable vs $24 on Opus. Fable's 3D worlds are better; motion graphics and landing pages come out like for like; neither writes natural marketing copy.

“The NEW Claude Fable 5.2 - No Hype Tests and 7 GREAT Use Cases” · Open the source


Claims ✓0 ~0 ✗0 ?53 reports
First stamped 10 Nov 2026Last re-stamped 13 Nov 2026