Mythical Anthropic AI Brings the Power of Its Legendary Model Unleashed to Common Users
Anthropic recently announced Fable, the start of a new family of models that brings many of the capabilities of its Mythos system to the public. As a refresher, the Mythos is a modern Anthropic model released in early April by Project Glasswing. The project saw Anthropic share access to the model with selected partners, including Apple and NVIDIA, with the aim of helping those organizations harden their software against AI cyberattacks. Glasswing also pushed the White House to rethink its policy on AI regulation. Last week, Anthropic said it was working very quickly to release a capable model of the Mythos with “strong protections” to the public, and today it’s doing just that with the Fable.
Anthropic says the new system, which starts in version five to align with the company’s other models, offers capabilities that “exceed” any that have been made publicly available in the past. “When the job is long and complex, the Fable 5 leads the way over our other models,” said Anthropic. In the company’s own testing, Fable beat not only its previous offering, Opus 4.8, but also GPT-5.5 and Gemini 3.1 Pro from rivals OpenAI and Google. Anthropic adds that Fable is particularly adept at tasks such as software engineering and document analysis, but also excels in domains where the company’s past models have typically underperformed. Vision is one such place.
You may remember last year Anthropic aired Claude trying (and failing) to hit Pokémon Red. The Anthropic model used for that Twitch series, 3.7 Sonnet, required an overlay for Claude to make sense of the game’s pixel art interface and world. If you watched the stream closely, you would notice that each time Claude moved a game character he had to re-evaluate his position afterwards. Between frames, Claude was blind, and couldn’t “hear” when the player character hit an obstacle like a tree. Conversely, Fable can beat Red Redthe 2004 Game Boy Advance remake, “with a small harness, with only a view.” Essentially, the company says that Fable’s new capabilities translate to a model that “can extract precise numbers from detailed scientific calculations and can perform complex vision-based tasks such as reconstructing the source code of a web application from screenshots alone.”
As for those protections I mentioned earlier, Claude will automatically transfer commands to other titles with Opus 4.8 powerless. “Releasing [Fable 5] safely and quickly, we carefully trigger these protections – sometimes they will catch harmless requests, although they start, on average, less than 5 percent of the time,” the company said.
On Tuesday, Anthropic also released the Mythos 5, which uses the same basic model as the Fable 5 but comes with a few protections. The company will initially offer the Mythos 5 through Project Glasswing before making it more widely available through a trusted access program. Claude’s subscribers can try Fable 5 until June 22. After that, using Fable 5 will cost usage credits, Anthropic with a value of 10 dollars per million input tokens and $50 per million withdrawal tokens. “After this point – when sufficient capacity allows us to do so – we intend to return Fable 5 as a regular part of the subscription programs. We intend to do this as soon as possible.”
Correction 5:24PM ET. A previous version of this article incorrectly stated that Legend 5 could design drugs. That ability is limited to Mythos 5. We apologize for the error.
