- 3 min read
- 5 key claims
- 2 sources cited
- For Developers, Businesses, AI enthusiasts
Overview
Simon Willison reports that Anthropic released Fable 5.2 and Mythos 5.2 and added Fable 5.2 to the llm-anthropic tool for running prompts [1]. The same report says Fable 5.2 is currently ranked first on Artificial Analysis, without specifying the ranking category or methodology [1]. Anthropic's statement that Fable 5.2 "solved two open problems in the physics of black holes" is recorded here as a vendor claim and has not been independently verified [1]. The available material does not describe Mythos 5.2.
Why it matters
Developers who use Anthropic-specific tooling can already route prompts to Fable 5.2 through llm-anthropic [1]. Businesses now have two additional names in Anthropic's model range to evaluate, but the available material provides no task-level comparison showing where either model differs from Opus 5.5 or from competing providers [1][2].
Opportunities
The documented access path is a compact installation and prompt sequence:
uv tool install llm
llm install llm-anthropic
llm keys set anthropic
llm -m claude-fable-5.2 "Generate an SVG of a pelican riding a bicycle"
One author used this command at all five reasoning levels to produce a pelican-riding-a-bicycle SVG [1]. This is a demonstration of prompt execution, not a quality or reasoning benchmark: the excerpt names only three of the five level labels — low, medium and high — and reports no measurement of reasoning quality. For evaluation work, the reported top position on Artificial Analysis gives Fable 5.2 a lead worth examining against a user's own workloads [1].
Risks
Adopting a newly released model on the basis of an unverified capability statement or an unspecified ranking carries two distinct risks. The black-hole physics claim originates with Anthropic and has not been checked by an independent party [1]. Separately, the reported Artificial Analysis position is not tied to a stated category or methodology in the available report, so it does not by itself establish performance on a particular workload [1]. The suggestion that these models approach or cross into AGI is speculation and should not guide procurement [2].
Context
Ethan Mollick reports having used Claude Fable 5.2 before its public launch and describes it as very impressive, following the earlier Opus 5.5 model [2]. That is one author's reported experience rather than a comparative evaluation. Mollick also frames the releases as raising a question about whether Anthropic's newest models cross into Artificial General Intelligence, while mentioning expected competing releases from OpenAI and Google. This is a debated question, not a result and not evidence that either Fable 5.2 or Mythos 5.2 is an AGI [2].
Known, claimed, inferred
The release extends Anthropic's lineup alongside the earlier Opus 5.5 line, and one technical writer with early access described Fable 5.2 as very impressive [2]. What can be established from the two reports is narrow: the two model names, the addition of Fable 5.2 to llm-anthropic, a reported first-place Artificial Analysis position without stated category or methodology, and an unverified Anthropic claim about black-hole physics [1]. Mythos 5.2 appears only as a released name; the reports say nothing about its purpose, capabilities or position [1][2]. The practical significance therefore rests on how Fable 5.2 behaves on a given organization's tasks rather than on the announcement alone.
Limitations & uncertainties
This page relies on two secondary reports: Simon Willison's Weblog and One Useful Thing [1][2]. The primary Anthropic release and the primary Artificial Analysis ranking record are not included; if they become available, they should be cited directly for the release details, the ranking category, its snapshot date and the benchmark methodology. The reports do not describe Mythos 5.2 beyond its name, and no independent third-party test report for Fable 5.2 or Mythos 5.2 is included [1][2]. Pricing, licensing, rate limits, context window and latency are not covered, and the available material gives no comparison between Fable 5.2 and Opus 5.5 beyond Mollick's qualitative assessment [1][2]. The black-hole physics claim remains unverified [1].
Evidence & sources
The key statements of this analysis, labeled by type, with the sources they rely on. Trendoviz does not run product tests.
- Verified factAnthropic released Fable 5.2 and Mythos 5.2.[1]
- Verified factFable 5.2 was added to the llm-anthropic tool for running prompts.[1]
- Primary-source or vendor statementAnthropic says Fable 5.2 solved two open problems in the physics of black holes.[1]
- User testimonialOne author used Claude Fable 5.2 before its public launch and described it as very impressive.[2]
- HypothesisMollick asks whether Anthropic's newest models cross into Artificial General Intelligence and mentions expected competing releases from OpenAI and Google; this is a question, not a finding.[2]
Sources
- 1llm-anthropic 0.31simonwillison.netretrieved Nov 18, 2026
- 2Fable 5.2 and Jagged AGIoneusefulthing.orgretrieved Nov 18, 2026
Updates & corrections
- First published · approved by an editor
Found an error? Report it.