The Mercury 3 model has expanded by more than 40 times, still fast in generation speed, with the price of one million output tokens as low as $0.35.

The Mercury 3 model has expanded by more than 40 times to 175B parameters, with an 8-fold increase in the speed of generating one million output tokens, and the price dropped to $0.35. This diffusion LLM scored 66.3 in the MMLU evaluation, surpassing GPT-6 Luna and starting to challenge mainstream large models.