It's a bizarre twist in the AI race! Anthropic's Fable 5, a highly advanced model, was recently hailed as the most capable AI publicly available, outperforming OpenAI's GPT 5.5 across a range of key benchmarks. For a fleeting three days, Fable 5 dominated leaderboards, including the Chatbot Arena, and showed significant leads in coding tasks, surpassing GPT 5.5 by substantial margins on benchmarks like SWE-Bench Pro and FrontierCode Diamond. It even offered unprecedented reasoning capabilities to its subscribers.

However, this reign was short-lived. On June 12th, the US government intervened, ordering Anthropic to pull Fable 5 offline. The stated reason was a jailbreak vulnerability. This abrupt shutdown has repositioned GPT 5.5 as the leading AI model that users and developers can actually access, not due to its own advancement, but because its primary competitor was removed from the market. While GPT 5.5 does show some strength in interactive, real-time coding tasks (Terminal-Bench 2.0), Fable 5's performance gap in broader software engineering and complex coding challenges was significant, often representing a 20-point difference or more on key metrics.
Adding to the complexity, OpenAI's GPT 5.5 is also more cost-effective, priced at half that of Fable 5. This makes it a more practical choice for high-volume applications where cost is a critical factor. The sudden removal of Fable 5 means that developers who were planning to integrate it into their workflows have had to revert to GPT 5.5 or older Anthropic models, experiencing a noticeable downgrade in capability, especially for coding-intensive tasks. The future of Fable 5 now hinges on Anthropic's negotiations with the government regarding its export control classification, with the company arguing that the identified vulnerabilities are minor and publicly known.
Fuente Original: https://thenextweb.com/news/anthropic-fable-5-vs-openai-gpt-5-5-benchmark-comparison
Artículos relacionados de LaRebelión:
- IA de Busqueda Revolucionaria Harness-1 Supera a GPT-54
- MiniMax M3 GPT-55 Gemini Outperformed at Fraction of Cost
- Anthropics AI Agents Now Learn From Mistakes
- Poolside Launches Free Open-Source Coding AI Model
- OpenAI Unveils GPT-55 Smarter Coding and Efficiency
Artículo generado mediante LaRebelionBOT












