Artificial Analysis Leaderboard: Claude Fable 5 and GPT-5.6 Sol Top Intelligence

Artificial Analysis released an updated LLM leaderboard ranking over 250 models from providers like OpenAI, Anthropic, Google, and SpaceXAI across five key metrics: intelligence, price, output speed, latency, and context window. For raw intelligence, Claude Fable 5 (with fallback) leads at 60, closely followed by GPT-5.6 Sol (max) at 59. In the speed category, Mercury 2 and Granite 4.0 H Small are the fastest models in terms of tokens per second, while North Mini Code and Gemini 2.5 Flash-Lite have the lowest first-token latency. On the cost front, Gemma 3n E4B and Nova Micro are the cheapest options. For context-sensitive tasks, Llama 4 Scout and Grok 4.20 0309 support the largest context windows, enabling processing of longer documents.

The leaderboard also provides a detailed table with blended pricing, median tokens per second, latency, and total response times. Notably, models like GPT-5.6 Terra (max) offer a strong balance of speed (136 tokens/s) and intelligence (55) at a lower price of $2.17 per million tokens. DeepSeek V4 Flash (max) is the most cost-effective at $0.06 per million tokens with a 40 intelligence score. The breakdown by reasoning mode (non-reasoning, low, medium, high, xhigh, max) allows developers to fine-tune their selection based on whether they need quick responses or deeper cognitive processing. This comprehensive comparison is crucial as businesses increasingly rely on LLMs for different use cases, from real-time chatbots to complex analysis.