Anthropic Unveils Fable 5: A Safer Mythos That Redefines Performance, But Comes with Caveats

Fazt
#anthropic #Fable 5 #llm #ai safety #cost model

Anthropic has officially launched Fable 5, its most potent large language model, now accessible to the general public. Positioned as the successor to Claude Opus 4.8, Fable 5 is a public-ready iteration of the previously restricted “Mythos” model, which was withheld due to concerns over its potential for identifying code vulnerabilities, executing cyberattacks, and even synthesizing bio-weapons. Anthropic states that Fable 5 has undergone rigorous cybersecurity enhancements, including Red Team testing, demonstrating its ability to avoid cyber-offensive activities, generating malicious code, or facilitating the creation of harmful biological or chemical agents. Benchmarks place Fable 5 as the leading model, outperforming competitors like GPT and Gemini, particularly in agent code tasks (SW Bench Pro). Its enhanced vision capabilities allow for processing complex images to generate highly accurate websites, animations, and even demonstrate sophisticated game-playing (e.g., Pokémon) and 3D modeling. To mitigate concerns about model distillation, Anthropic has implemented a system that detects attempts to train other models and automatically switches the user to Claude Opus 4.8.

While Fable 5 is immediately available across Anthropic’s Claude platform, Cursor, GitHub, and via partners like Cognition Labs (creator of Devin), its operational cost presents a significant caveat. The model consumes tokens at double the rate of Claude Opus 4.8, with pricing set at $10 per million input tokens and $50 per million output tokens. For existing Max and Pro subscribers, Fable 5 will be available without additional charge only until June 22nd or 23rd, after which it will transition to a pay-per-use model. Anthropic attributes this temporary pricing structure to insufficient operational capacity, promising to reintegrate Fable 5 into subscription plans once infrastructure scales. This move has sparked user complaints regarding the rapid consumption of credits and high costs, despite widespread acknowledgment of the model’s superior quality and capabilities. The industry now watches closely for responses from major competitors like OpenAI as Fable 5 begins to integrate into broader daily use.