~/AI MODELS/claude-fable-5-2-scores-record-65-on-artificial-analysis

Claude Fable 5.2 Scores Record 65 on Artificial Analysis

Claude Fable 5.2 has achieved the highest intelligence score of 65 on Artificial Analysis, setting a new benchmark record. However, users report a substantial increase in token usage and cost per task compared to previous versions. This milestone highlights the ongoing push for higher intelligence in AI models, but it also underscores the growing trade-offs between computational performance and cost efficiency. Developers and companies must carefully evaluate whether the performance gains justify the steep rise in operational expenses. Data shows that output tokens per task nearly doubled from 118k in version 5.1 to 274k in version 5.2, driven largely by an explosion in reasoning tokens. Despite these cost concerns, the model performs competitively on the Pareto frontier for speed and intelligence.

## BACKGROUND

Artificial Analysis is an independent platform that benchmarks AI models across key performance metrics such as quality, pricing, and latency. Intelligence scores and Pareto efficiency frontiers help developers understand the trade-offs between how fast and how smart an AI model is.

## REFERENCES

## KEYWORDS

#AI Models#LLM Benchmarks#Claude#Artificial Analysis#Cost Efficiency

$ subscribe --daily