Mercury 3 Is INSANE: I Tested 13 Live Examples (+ 100 MILLION FREE Tokens)

I tried Inception Labs' Mercury 3, a diffusion language model, on 13 live examples (games, tools, a landing page, a puzzle, translation, a search trap) and then timed it through the API. New accounts get 100 million free tokens (per the docs). The blog claims 1,068 tokens/sec; in my 3 runs I measured 258-416 tokens/sec end to end, and I explain why. Try it: https://chat.inceptionlabs.ai/ | Docs: https://docs.inceptionlabs.ai/get-started | API keys: https://platform.inceptionlabs.ai/dashboard/api-keys Launch post: https://www.inceptionlabs.ai/blog/introducing-mercury-3 Cheap, reliable GPUs: Run