Quick comparison of the leaked Llama 3.1 benchmarks. Across the 8 common ones Claude 3.5 Sonnet stays top with an average of 82.8%, 405B slots in just behind at 81.4% beating GPT-4o into bronze position with 80.9%...

@AnthropicAI have really aggressively priced #Claude3 Haiku and the buzz, plus position above the curve suggest... it is now the most attractive model option. @OpenAI under pressure at both ends of the market.