Fable
An Anthropic model derived from the Mythos tier, with capabilities reduced relative to the original Mythos specification.
Core Facts
Claude is Anthropic's AI large language model.
In the deep audit benchmark, Sonnet 5.5's cost is approximately half that of Opus, its score is slightly higher than Opus, and it takes about 5 minutes (roughly half the time Opus takes).
Accounting for real usage scenarios, the actual cost advantage of Opus 5 over Fable is approximately 20% to 25%, with Fable costing about $2.75 per task and Opus 5 about $2.03.
Timeline (last 90 days)
On MemoryBench, Wonder defeated Fable, Opus, and Astra; on the Artificial Analysis index, Wonder Full ranks second, behind only Opus 5.5.
In the creator's self-made code audit benchmark, the Full version performed close to Astra and Fable, while the base version Muse Wonder scored almost only half of the Full version, a gap of more than 30%.
Theo said on a livestream himself that buying Fable-level intelligence at the price of GLM 5.3 Flash 'will absolutely happen'.
On the DeepSuite benchmark, Muse Wonder topped the list with 79 points, the smaller version scored 78.8, Fable scored 67, Astra scored 71, and Opus 5.5 scored 74.
On Terminal Bench, the two Muse models scored 57.1 and 52 respectively, roughly on par with the 52 to 59 range of Fable, Astra, and Opus.
In terms of per-task cost, Muse Wonder Max is $4.52, Sonnet, Opus, and Fable have the highest per-task cost at $7.63, and GPT 6.1 Sol Max is only $0.72 per task.
In Meta's internal memory tests, Muse Wonder outperformed systems such as Fable, Opus, and Astra.
Fable level refers to a tier of model standard previously considered within Anthropic to be at a higher capability level.
Anthropic's 'Teamer' mode is coordinated by a strong model, with cheap models executing batch work; in one test, the total cost of Fable leading 25 Sonnet workers was about half that of using Fable alone, with time reduced from more than 15 hours to less than 2.5 hours, and the quality score 10 to 12 points lower.
Fable focuses more on outputting content at the level of a 'complete work' rather than fragmentary text completion.
40 more timeline events
All Facts (20)
Claude is Anthropic's AI large language model.
70%VerifiedIn the deep audit benchmark, Sonnet 5.5's cost is approximately half that of Opus, its score is slightly higher than Opus, and it takes about 5 minutes (roughly half the time Opus takes).
65%VerifiedAccounting for real usage scenarios, the actual cost advantage of Opus 5 over Fable is approximately 20% to 25%, with Fable costing about $2.75 per task and Opus 5 about $2.03.
65%VerifiedOpenAI's naming system shifted during the GPT-4 era from decimal version numbers to letter suffixes (such as 4-turbo and 4o) to identify architectural optimizations.
65%Partially VerifiedSonnet 5.5 is on par with Opus 5.5 in offensive and defensive cybersecurity capabilities, making it the first model in the Sonnet series to adopt equivalent cybersecurity protection mechanisms.
65%UnverifiedOn MemoryBench, Wonder defeated Fable, Opus, and Astra; on the Artificial Analysis index, Wonder Full ranks second, behind only Opus 5.5.
50%UnverifiedIn the creator's self-made code audit benchmark, the Full version performed close to Astra and Fable, while the base version Muse Wonder scored almost only half of the Full version, a gap of more than 30%.
50%UnverifiedIn terms of per-task cost, Muse Wonder Max is $4.52, Sonnet, Opus, and Fable have the highest per-task cost at $7.63, and GPT 6.1 Sol Max is only $0.72 per task.
50%UnverifiedOn the DeepSuite benchmark, Muse Wonder topped the list with 79 points, the smaller version scored 78.8, Fable scored 67, Astra scored 71, and Opus 5.5 scored 74.
50%UnverifiedOn Terminal Bench, the two Muse models scored 57.1 and 52 respectively, roughly on par with the 52 to 59 range of Fable, Astra, and Opus.
50%UnverifiedIn Meta's internal memory tests, Muse Wonder outperformed systems such as Fable, Opus, and Astra.
50%UnverifiedFable level refers to a tier of model standard previously considered within Anthropic to be at a higher capability level.
50%UnverifiedAnthropic's 'Teamer' mode is coordinated by a strong model, with cheap models executing batch work; in one test, the total cost of Fable leading 25 Sonnet workers was about half that of using Fable alone, with time reduced from more than 15 hours to less than 2.5 hours, and the quality score 10 to 12 points lower.
50%UnverifiedFable is an AI-driven creativity and content generation tool that focuses on helping users quickly generate structured narrative content, including story creation, brand copywriting, and interactive experience design.
50%UnverifiedFable focuses more on outputting content at the level of a 'complete work' rather than fragmentary text completion.
50%UnverifiedIn a demonstration classifying 100 emails (hot leads/cold leads/non-leads), Jev completed all classifications in under one second, outperforming Haiku and Fable in both speed and cost.
50%UnverifiedFable is believed to specialize in code generation and visual rendering tasks, including 3D game development, low-level shader rendering, and front-end animation effects.
50%UnverifiedOn the Frontier Code reasoning benchmark, the medium reasoning tier performed better than the high and extra high tiers, approaching the max tier; this phenomenon occurred with both Opus and Fable.
50%UnverifiedSome Reddit users speculated that the biggest outcome of Project Glasswing may have been to allow Anthropic to train its subsequent model Fable on closed-source code that it otherwise would not have been able to access.
50%UnverifiedThe author recommends that Sonnet 5.5 should not be called directly, but rather used as a sub-agent component orchestrated by Opus or Fable when decomposing complex tasks.
50%