Claude Fable 5.1 tops Artificial Analysis Index but costs 20% more (twitter.com)

Anthropic has launched Claude Fable 5.1, which has achieved a remarkable score of 66 on the Artificial Analysis Intelligence Index, surpassing all previous models, including Claude Opus 5 and GPT-5.6 Sol. This new model shows significant improvements across various benchmarks, notably in high-level evaluation (HLE) with a score of 59.1%, and it excels in both the Terminal-Bench v2.1 and SciCode evaluations. Despite these advancements, Fable 5.1 comes with a 20% higher cost per task compared to its predecessor, Fable 5. The increased expense can be attributed to its use of approximately 1.7 times more output tokens, although a 75% reduction in cache read prices — from $1 to $0.25 per million cached tokens — helps to offset some of the costs. The launch of Claude Fable 5.1 is significant for the AI/ML community as it showcases the continuous evolution of large language models and their performance in complex tasks. Its extended context window of 1 million tokens and capability to process both image and text inputs highlight its versatility. While it leads in agentic work evaluations, demonstrating superior analytical quality, it effectively ties with Claude Opus 5 in other aspects, reflecting a competitive landscape in AI model development. As developers and organizations assess the cost-benefit dynamics of such advanced models, Fable 5.1 sets a new benchmark for balancing intelligence with operational cost efficiency.