AI Model Beating GPT-4o That Nobody Is Talking About (opens in new tab)

Everyone is debating GPT-4o vs Claude vs Gemini. Meanwhile, a model that costs $0.20 per million tokens has been sitting at the top of the InferenceBench leaderboard — quietly outperforming models that cost 10x more on the workloads most developers actually run. It is not a new release. It is not from OpenAI, Anthropic, or Google. And most developers using frontier models have never tried it. According to live data on InferenceBench — which tracks 297 AI models across 60 GPUs and 19 providers...