Anthropic's Claude Fable 5.2 raises the bar for technical work, AWS says
In a demonstration, 1,000 Fable 5.2 agents formalized the proof of Fermat’s Last Theorem in 30 hours, TNW reported.
Rhea Lease · AI writer ·
1 primary · 4 independent · 4 commentary · 1 vendorA crowd of agents assembles a mathematical proof while some research uses remain restricted.Image courtesy of Fay Tograph · AI photographer, made with ChatGPT image generation
Claude Fable 5.2 is Anthropic’s most capable generally available model for technical work that raises the bar in coding and the sciences, according to Amazon Web Services.[1]
If the performance holds, researchers and engineers gain a tool that can formalize mathematical proofs, design hardware, and conduct long-running scientific analysis with less manual intervention, according to Anthropic.[1][2]
The models share the same core capabilities, but Fable 5.2 blocks certain biology and cybersecurity use cases, while Mythos 5.2 has more relaxed guardrails and is only available to trusted researchers through a pair of limited-access programs, SiliconANGLE reported.[2]In one demonstration, a team of 1,000 Fable 5.2 agents formalized the proof of Fermat’s Last Theorem in 30 hours, a task that previously took 11 days, TNW reported.[3]
Fable 5.2 costs $8 per million input tokens, according to TNW.[3]An optional high-performance mode generates prompt responses 2.5 times faster than the standard tier for double the price, SiliconANGLE reported.[2]
What people on X are saying, summarized by Grok
People on X are split. Some treat the Fermat run as proof that large agent teams can stay on a long technical job, and a few set it against a human project expected to take years. Others say the agents translated a known proof into a form a computer can check, rather than discovering new math, and that the result is the company's own single demo. Several add that the pricey model only looks worth it on the longest jobs, while a minority shrugs that the demo is less useful than everyday work.
Posts chosen and quoted by Grok
Everyone is hyping the Fable 5.2 "1000 agents" Fermat proof, but they're missing the actual story.
The formalized proof is machine-checkable, which is incredible. But it's not a new mathematical discovery,it's a translation of an existing human proof into Lean.
The real headline? Anthropic just proved they can coordinate 1,000 AI agents on a single long-horizon task without them hallucinating at …
Anthropic shipped its new #1 model today, but its own launch post tells you not to start with it??
for most coding and knowledge work, the recomment starting with opus 5.5, "our strongest daily driver".
i wanted to know why
1. Everyday agent work first.
- on osworld, where the model drives a real desktop and clicks through apps to finish a task, fable gets 89% and opus 87%.
- on code a …
In early 2024, a top mathematician named Kevin Buzzard launched a major human project (backed by a multi-million-dollar grant) to formalize Fermat's Last Theorem in the Lean programming language, estimating it would take a global team of math PhDs 5 years to code. Now here we go.
Written by Grok from posts on X, with post text as Grok returned it. Not reported or checked by us, and not checked against X. The story above does not rely on it.