The diffusion-model lab's newest release ships with a one-time token grant and a 1M context window.
Mercury 3 is Inception Labs' newest diffusion language model family — Mercury, Mercury Voice, and Mercury Decide — with a 1M token context window (up from 260K on Mercury 2.5) and speeds over 1,000 tokens per second on NVIDIA GPUs. The company claims it outperforms GPT-6 Luna and Claude Haiku 5.5.
The catch: 100M tokens are a one-time grant per account; no recurring free tier. Standard pricing is $0.20/M input and $0.70/M output, with a 50% launch promo at $0.10/M and $0.35/M. Cached tokens cost 90% less.
How to claim: 1) Sign up at https://platform.inceptionlabs.ai/ → 2) Create an API key → 3) 100M tokens apply automatically.
Deadline: Ongoing.
Link: Introducing Mercury 3