<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/RMKtq3Tup" -->

---
title: Claude Fable 5.2 and Mythos 5.2 land at $8/$40 | daily.dev
description: Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, priced on AWS Bedrock at $8 per million input tokens and $40 per million output tokens. Earlier...
canonical: https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Claude Fable 5.2 and Mythos 5.2 land at $8/$40 | daily.dev
og:description: Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, priced on AWS Bedrock at $8 per million input tokens and $40 per million output tokens. Earlier...
og:url: https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup
og:image: https://api.daily.dev/og/posts/RMKtq3Tup.png
og:image:alt: Claude Fable 5.2 and Mythos 5.2 land at $8/$40
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude Fable 5.2 and Mythos 5.2 land at $8/$40

**[Trends](https://daily.dev/sources/trends)** · 3 min read · 12 upvotes · 3 comments

## Summary

Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, priced on AWS Bedrock at $8 per million input tokens and $40 per million output tokens. Earlier traces had pointed to a version 5.5, raising questions about whether 5.2 is a smaller upgrade than expected. Claude Fable 5.2 is already live in v0, while Anthropic's own announcement just calls them 'available now' without further detail. The pricing is drawing attention since output tokens drive costs in long coding and tool-use agent loops.

## Content

The leaderboard result is clean: Claude Fable 5.2 sits at 65 on Artificial Analysis, five points clear of Gemini 4 Carbon. The advice coming back from users, and from Anthropic itself, is just as consistent: don't make it your default.

**Anthropic's own people say start on Opus.** @addyosmani: "I'd still start most work on Opus 5.5. It's our strongest daily driver." @felixrieseberg: Fable "thinks a little slower" and wins on big, long jobs like codebase rewrites. The pairing that keeps showing up is Fable for planning, Opus for execution. @mattshumer_ calls it "the new meta." @AlexFinn, who says he tested it for months, calls Fable "the most intelligent model ever created" and wants it to write the tasks while Opus does the work.

**The cost argument is the real fight.** Fable is $8/$40 per million tokens, double Opus, though 20% below Fable 5.1. Vendors say per-task cost is what counts: Fable at high effort roughly matches Opus at max for about half the price. Critics point at the other end of the dial. At max effort it costs $26.54 per task on Artificial Analysis and takes 45 minutes. @nssntus: "use the top model" is the wrong default now, since first and third place on MKA are 4 points apart. @overshadowed_x notes Gemini 4 Carbon runs $7.76 a task versus Fable's $9.08 at high effort. Model ML found Fable at 2 to 2.6x Opus's per-task cost.

**@theo is the grumpy voice.** He says Fable burns usage faster per dollar than Opus, and is capped at 50% of the weekly limit, so using it until it runs dry and then switching costs you $4k a month. His $42 Fishslop run was "not significantly better than Opus 5.5." His take on the OS demo: "yeah, and?" Then he posts the Artificial Analysis chart and says Fable is "mogging so hard."

**@omarsar0 is tired of the demos.** He calls the early-tester posts "crappy games and the same old stuff," and wants people to try test-time compute on scientific discovery.

**Mixed results elsewhere.** A reviewer's KingBench 4 run gave Fable 70% to Carbon's 53.3%, at $222 versus $23. But Carbon's summaries overclaimed broken features. Fable still loses to GPT-6 Astra at Blood on the Clocktower (@dwarkesh_sp via a repost) and on the weight-space Breakout test, 493 vs 517.

**The headline claim.** 1,000 agents formalized Fermat's Last Theorem in about 30 hours, versus 11 days for the earlier attempt. Anthropic also says a Fable agent team built a working OS over 8 days.

**Safety footnote.** Mythos 5.2, the same model with looser cyber safeguards for vetted orgs, used a planted sandbox escape in 3.1% of scenarios, down from 21% for 5.1. Anthropic says it also seems to suspect it's being evaluated more often, up to 43% of audit transcripts. @antirez, on early Mythos access, found two severe bugs in 20 minutes: "Mythos is good."

## Questions this post answers

### How much does Claude Fable 5.2 cost per million tokens on AWS Bedrock?

Claude Fable 5.2 and Claude Mythos 5.2 are priced at $8 per million input tokens and $40 per million output tokens on AWS Bedrock. Output pricing matters most for agent builders since long-running coding and tool-use loops burn through output tokens quickly, making a chatty agent expensive fast.

_Comparing model costs for agent workloads is easier when pricing changes land on daily.dev as they happen._

### Is Claude 5.2 the same model that leaked as version 5.5?

It's unclear. Testing accounts spotted traces pointing to a version 5.5 before launch, but Anthropic shipped Claude Fable and Mythos as 5.2 instead, leaving open whether the number changed or the earlier traces were simply wrong. Hands-on comparisons are needed to tell if 5.2 is a smaller step than the leaked numbering suggested.

_Developers tracking which Claude version actually ships can follow the comparisons as they surface on daily.dev._

## Community take

How the wider developer community reacted, aggregated from 4 discussions and 677 comments across hackernews, x (as of 2026-11-16).

**TL;DR:** The discussion is dominated by off-topic crypto hype replies, but the substantive thread centers on a workflow where Fable 5.2 interrogates users with many questions before building, splitting ideas into tracked issues, then handing execution to another model. Claude Fable 5.2 and Mythos 5.2: Most discussion is dominated by accusations that prior model versions were secretly degraded ('nerfed') ahead of this release, with heated debate over whether this is real, A/B testing, compute constraints, or perception bias; the headline benchmark achievements (Fermat formalization, physics results) get comparatively little engagement. Fable 5.2 one-shotted my nephew's dream game: Reactions split between delight at a kid's drawings becoming a real playable game and unease about AI's effect on children's creative development, learning, and the generic 'AI slop' aesthetic of the generated art and animation.

**Sentiment:** 21% positive · 36% mixed · 43% skeptical

**The case for**

- The upfront question-asking phase surfaces edge cases and assumptions before code is written, reducing rework.
- Some report the new/updated model solving tasks faster with no noticeable quality drop in daily coding use.
- Several see this as a positive, low-stakes way for kids to realize their creative visions quickly rather than being gatekept by years of learning to draw or code.
- At medium/high effort settings, it's considered highly cost-competitive and near the pareto frontier of cost vs. intelligence.
- Saving the Q&A as a spec file lets future sessions reload context instead of re-asking the same questions.

**The pushback**

- Asking 100+ questions upfront can eat into usage limits before any code is written.
- Many suspect models are deliberately nerfed/quantized before new version releases to boost comparative benchmarks or save compute.
- Multiple commenters worry this could stunt development of drawing, coding, patience, and the satisfaction of working through difficulty.
- Running at default 'max' effort massively inflates token usage and cost per task compared to Fable 5.1.
- Early answers are sometimes just guesses that change once the first working build appears.

**By community**

- hackernews (heated): Discussion splits sharply between people convinced Anthropic secretly degrades models before releases and others dismissing this as unfounded pattern-seeking or placebo effect, with a smaller side debate on writing style and AI's broader societal trajectory.
- x (mixed): Most engagement is low-signal hype/spam unrelated to the model, while a smaller substantive thread debates the merits and pitfalls of a question-heavy planning workflow with Fable 5.2.

**Hottest debate:** Whether the extensive upfront-questioning workflow is a genuine productivity unlock or just a usage-limit-draining repeat of existing techniques like 'grill-me' skills.

**Open questions**

- Does a detailed upfront plan stay useful once the actual build starts drifting from it?
- Is there concrete, reproducible evidence of model nerfing versus natural variance/A/B testing?
- Will kids who grow up with AI-generated creative output still develop curiosity about how things work under the hood?
- Where exactly is the line between users who truly need frontier-level models versus those over-provisioning for simple tasks?
- How should an agent handle a question that arises mid-build after the initial Q&A phase is done?

**Highlights**

> @AlexFinn Making the agent interrogate you first makes sense. My worry is 100+ questions eating the weekly limit before any code gets written.
> — [honza\_vojacek on x · 1 points](https://x.com/honza_vojacek/status/2122389832898895966)

> > Our first formalization of Fermat’s Last Theorem, performed by an unreleased research version of Claude, took 11 days. It relied on a purpose-built harness along with guidance from a human researcher. By contrast, a team of 1,000 Fable 5.2 agents fully formalized the theorem in just 30 hours, with minimal human intervention. Fable 5.2 started from minimal prompts, built its own ways to coordinate, and used no mathematical references, only a list of intermediate theorem statements to score progress. It needed one-third of the action tokens as our first formalization and produced a proof seven times shorter. As can be seen in the figure below, Opus 5.5 (with the same 1,000-agent setup) completed only 26% of the formalization in 18 hours, and barely made any more progress after 30. Most people don't really have intuitions on what it means to push the frontier now (because AI has surpassed everything they are good at, for example, coding). We are truly inside the singularity (using Kurzweil's definition, slightly paraphrased - the time when technology progresses so fast no one can keep track with it).
> — [yewenjie on hackernews · 9 comments](https://news.ycombinator.com/item?id=50430379)

> > a team of 1,000 Fable 5.2 agents Genuinely curious what that costs in terms of energy input and amortized computer cost vs. the previous version + a human's salary for 11 days.
> — [gwerbin on hackernews · 1 comments](https://news.ycombinator.com/item?id=50431080)

> @AlexFinn 100 questions is fine. I'd read the 40 tickets before the agent starts: one wrong answer at question 12 ends up in half of them.
> — [Moretti\_ale90 on x](https://x.com/Moretti_ale90/status/2122400174404202984)

> @AlexFinn The workflow moves the bottleneck from planning to verification. One hundred questions and forty Linear issues give the agent a detailed map, but each ticket still needs to be checked against the ones that came before it. Context drift compounds when the agent works through them
> — [DAssetBuzz on x](https://x.com/DAssetBuzz/status/2122385606533607725)

**Source threads**

- [hackernews](https://news.ycombinator.com/item?id=50429906) · 580 points · 338 comments
- [hackernews](https://news.ycombinator.com/item?id=50433659) · 156 points · 124 comments
- [hackernews](https://news.ycombinator.com/item?id=50431144) · 178 points · 117 comments
- [x](https://x.com/AlexFinn/status/2122381648679608354) · 0 points · 98 comments
- [hackernews](https://news.ycombinator.com/item?id=50429068) · 1 points · 0 comments
- [x](https://x.com/stolinski/status/2122393711023927450) · 0 points · 0 comments

## Community discussion

Top comments from developers on daily.dev.

**@petecapecod** · 0 upvotes

> oh yeah the new big boy is out! Remember those Fable resets only reset Fable and not your weeklys

**@ernestoporto** · 0 upvotes

> People forget that it's expensive on the bleeding edge of technology

**@exeboii** · 0 upvotes

> AI is becoming too Expensive now, and we still Don't know what type of code it will produce because it is based on probabilities rather Than absolute Accuracy

---

Tags: [#llm](https://daily.dev/tags/llm), [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic), [#v0](https://daily.dev/tags/v0)

[View this post on daily.dev](https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"Claude Fable 5.2 and Mythos 5.2 land at $8/$40","url":"https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup"},"datePublished":"2026-11-09T19:00:41.650Z","dateModified":"2026-11-16T19:52:07.830Z","description":"Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, priced on AWS Bedrock at $8 per million input tokens and $40 per million output tokens. Earlier...","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":3,"discussionUrl":"https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":12},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":3}],"keywords":"llm,claude,anthropic,v0","timeRequired":"PT3M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Claude Fable 5.2 and Mythos 5.2 land at $8/$40"}]}
{"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup","comment":[{"@type":"Comment","text":"oh yeah the new big boy is out! Remember those Fable resets only reset Fable and not your weeklys","datePublished":"2026-11-10T13:57:30.847Z","url":"https://daily.dev/posts/RMKtq3Tup#c-q4URdTEU0","author":{"@type":"Person","name":"Peter Cruckshank","url":"https://daily.dev/petecapecod","image":"https://media.daily.dev/image/upload/s--ZJhQyKws--/f_auto/v1721235024/avatars/avatar_A9xh33q0QoxtkGoJRCosp"}},{"@type":"Comment","text":"People forget that it’s expensive on the bleeding edge of technology","datePublished":"2026-11-10T11:10:26.958Z","url":"https://daily.dev/posts/RMKtq3Tup#c-2q4dzYZ1b","author":{"@type":"Person","name":"Ernest Oporto","url":"https://daily.dev/ernestoporto","image":"https://media.daily.dev/image/upload/s--5At_l6_K--/f_auto/v1789389230/avatars/avatar_WlGqx4f45NvFnLVg0Bn2R?_a=BAMAMicg0"}},{"@type":"Comment","text":"AI is becoming too Expensive now, and we still Don’t know what type of code it will produce because it is based on probabilities rather Than absolute Accuracy","datePublished":"2026-11-10T04:45:10.537Z","url":"https://daily.dev/posts/RMKtq3Tup#c-VAgJVoNTZ","author":{"@type":"Person","name":"See Exe","url":"https://daily.dev/exeboii","image":"https://media.daily.dev/image/upload/s--v3BTnJ1t--/f_auto,pg_1/v1794289458/avatars/avatar_YFk50mGbwgETJbK796JwD?_a=BAMAMiiw0"}}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/claude-fable-5-2-and-mythos-5-2-land-at-8-40-rmktq3tup#faq","mainEntity":[{"@type":"Question","name":"How much does Claude Fable 5.2 cost per million tokens on AWS Bedrock?","acceptedAnswer":{"@type":"Answer","text":"Claude Fable 5.2 and Claude Mythos 5.2 are priced at $8 per million input tokens and $40 per million output tokens on AWS Bedrock. Output pricing matters most for agent builders since long-running coding and tool-use loops burn through output tokens quickly, making a chatty agent expensive fast. Comparing model costs for agent workloads is easier when pricing changes land on daily.dev as they happen."}},{"@type":"Question","name":"Is Claude 5.2 the same model that leaked as version 5.5?","acceptedAnswer":{"@type":"Answer","text":"It's unclear. Testing accounts spotted traces pointing to a version 5.5 before launch, but Anthropic shipped Claude Fable and Mythos as 5.2 instead, leaving open whether the number changed or the earlier traces were simply wrong. Hands-on comparisons are needed to tell if 5.2 is a smaller step than the leaked numbering suggested. Developers tracking which Claude version actually ships can follow the comparisons as they surface on daily.dev."}}]}
```

