Give Claude Fable 5.2 one paragraph describing a gadget, and Anthropic says it returns a circuit board ready for the factory. In Anthropic’s launch post for Fable and Mythos 5.2, the company says the model designed three working devices this way: a home voice-assistant hub, a USB speakerphone and a handheld virtual pet. All three worked without electrical fixes, in a test nobody outside has repeated yet.
Fable 5.2 and its restricted twin, Claude Mythos 5.2, came out on 9 November, as The Verge reported. They are Anthropic’s most capable models, and they cost less than the 5.1 versions.
What the model did in Anthropic’s tests
Anthropic says Claude wrote the specs, picked parts, laid out each circuit board, modeled the printed cases and wrote the software. Each board was ready in under a day, and the board maker Diode built them.
In another demo, it ran more than 1,000 subagents (helper copies that each take part of the job) for eight days and built a prototype operating system that boots on a laptop.
A team of 1,000 Fable 5.2 agents also translated the proof of Fermat’s Last Theorem into Lean, a language a computer can check line by line, in 30 hours. Anthropic’s earlier attempt took 11 days. Anthropic also says Fable 5.2 solved two open problems about black holes, written up in two preprints, papers not yet peer reviewed.
What outside tests found
All of that comes from the maker. The clearest outside check is Artificial Analysis, which runs one task set on every model. Its Claude Fable 5.2 results page ranks the model first of 96 on its Intelligence Index, with a score of 65. We later looked at where Google’s Gemini 4 Neon lands on that index.
That lead is slow and expensive. Fable 5.2 took 41.8 minutes per test task, against a median of 12.1 minutes for comparable models, and cost $26.54 per task. On the Artificial Analysis leaderboard, OpenAI’s GPT-6.2 Astra scores 62 for $4.00 per task. Our guide to the big AI leaderboards explains why one index never settles it.
Who can use it and what it costs
Anthropic’s Claude Fable product page lists Fable 5.2 on the Pro, Max, Ultra, Team and Enterprise plans. Developers get it through the API, AWS, Google Cloud and Microsoft Foundry.
The API price drops to $8 per million input tokens and $40 per million output tokens, from $10 and $50. Tokens are chunks of text; the Claude docs page for Fable 5.2 puts 1 million at roughly 555,000 words. That price equals what OpenAI charges for GPT-6.2 Astra. An optional fast mode runs up to 2.5 times faster for $16 and $80.
Anthropic tells most people to start with Claude Opus 5.5, at $4 and $20, and move up to Fable for the hardest open-ended work. Dianne Penn, Anthropic’s head of product management for research, told CNBC the model is meant for “power users or power use cases.”
Fable and Mythos: one model, two sets of locks
Fable and Mythos are the same underlying model. Fable has filters that block most security testing and research biology tasks and hand them to older models. Anthropic says Fable 5.2 blocked about 55% fewer requests than 5.1 in early testing. Mythos 5.2 runs without some of those filters and goes only to vetted security and life-science groups through Anthropic’s verification programs.
Fable also keeps your data for 30 days by default for safety monitoring, worth knowing if your company bans retention.
What to do this week
If you pay for Claude, give Fable 5.2 one hard, open-ended job, such as a messy data cleanup or a product spec, and run the same prompt through Opus 5.5. Compare the result, the time and the usage each burned. Keep Fable for jobs where the difference shows.
Sources
Get the weekly email
The week's most useful AI news, a tutorial and an interview.





codigopromocionalinkabet
The promo codes here actually deliver real value without any hidden tricks. I redeemed my welcome bonus and watched it multiply during my first session. Everything feels transparent and the wagering requirements are super fair for newcomers. Head over to codigopromocionalinkabet to grab your own starter deal today