Skip to content
Models · Nov 11, 2026

Anthropic releases Claude Fable 5.2 and Mythos 5.2, claiming top LLM spot on Anthropic’s IPO roadshow day

Fable 5.2 tops independent leaderboards while Mythos 5.2 expands agentic capabilities; both arrive with pricing changes and subscription restrictions.

Trust76
HypeSome hype

1 source · cross-referenced

ShareXLinkedInEmail
TL;DR
  • Anthropic released Claude Fable 5.2 and Mythos 5.2 on the first day of its IPO roadshow, claiming both models surpass Opus 5.5 in most evaluated capabilities.
  • Fable 5.2 is reported as the new #1 model on Artificial Analysis’ Intelligence Index v5 with a score of 65 (max), ahead of Gemini 4 Carbon (60) and Opus 5.5 (59).
  • Pricing for Fable 5.2 is $8 per million input tokens and $40 per million output tokens, a 20% reduction from Fable 5.1, but per-task costs rose due to higher token usage and subagent activity.
  • Mythos 5.2 introduces expanded agentic workflows, including a reported 1,000-agent run formalizing Fermat’s Last Theorem in 30 hours and designing electronic devices from short prompts.

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, the first day of its IPO roadshow, positioning both models as surpassing Opus 5.5 in nearly all evaluated capabilities. The company described the releases as built for "the most daunting technical challenges," highlighting gains in coding, reasoning, and vision.

Fable 5.2 is reported to have retaken the #1 spot on Artificial Analysis’ Intelligence Index v5 with a score of 65 at the max effort level, outperforming Gemini 4 Carbon (60) and Opus 5.5 (59). Anthropic also claims Fable 5.2 at High effort matched Opus 5.5 at Max effort on FrontierCode 2.0 for about half the cost per task, while Fable 5.2 at Xhigh was the top released model on the same benchmark.

Pricing for Fable 5.2 is set at $8 per million input tokens and $40 per million output tokens, a 20% reduction from Fable 5.1’s $10/$50 pricing. However, SemiAnalysis and Artificial Analysis report higher per-task costs due to increased token usage and the model’s tendency to run hundreds or thousands of test cases and subagents during tasks. For example, SemiAnalysis observed a rise in per-task cost from $67.16 to $88.15 when comparing Fable 5.1 to 5.2 on GDPx, a benchmark of professional tasks.

Mythos 5.2 expands agentic capabilities, with Anthropic reporting that 1,000 Fable 5.2 agents formalized the existing proof of Fermat’s Last Theorem in 30 hours, compared to Opus 5.5 completing 26% of the proof in 18 hours under the same setup. Mythos 5.2 also accurately modeled 28 of 34 gene-therapy capsids. Long-horizon demos include an operating system built from an empty folder by over 1,000 subagents over 8 days, which boots on a laptop and includes its own kernel, drivers, persistent storage, and windowing system.

Subscription access to Fable 5.2 is tiered: Ultra and Team Premium users receive scaled "resets" based on tenure (1–3 resets), usable until March 8, while Pro and Team Standard users get a one-week trial capped at half the weekly limit plus one reset. Fable 5.2 can consume no more than 50% of a user’s weekly limit. Zero data retention is available only to eligible enterprise customers.

The models introduce five effort levels—low, medium, high, xhigh, and max—with medium as the default in Claude apps and Claude Code. Anthropic’s system card refers to a single shared base model underlying both Fable 5.2 and Mythos 5.2. The rollout includes integration with Cursor, GitHub Copilot/VS Code, Perplexity Computer (Pro/Max) and Search (Max), Factory, Hermes Agent via Nous Portal and OpenRouter, Arena, and DesignArena.

Sources
  1. 01Latent Space — swyx — [AINews] Fable 5.2 retakes World #1 LLM spot as $ANT IPO roadshow begins
Also on Models

Stories may contain errors. Dispatch is assembled with AI assistance and curated by human editors; despite the trust-score filter, mistakes happen. We correct publicly — every article links to its revision history. Nothing here is financial, legal, or medical advice. Verify before relying on any claim.

© 2026 Dispatch. No ads. No sponsorships. No paid placement. Reader-supported via Ko-fi.

Built by a person who cares about honest AI news.