Comment by kristianp

How big is Mercury 2? How many tokens is it trained on?

Is it's agentic accuracy good enough to operate, say, coding agents without needing a larger model to do more difficult tasks?