NousResearch/Hermes-4-14B

Hermes 4 14B by Nous Research is a 14 billion parameter, hybrid-mode reasoning model based on Qwen 3. It features a significantly expanded post-training corpus of approximately 5 million samples and 60 billion tokens, emphasizing verified reasoning traces. This model excels in math, code, STEM, logic, creativity, and schema adherence, while offering improved steerability and reduced refusal rates, making it highly adaptable for various aligned applications.