← All model drops
Live drop
Mercury 3
InceptionDropped Oct 19, 2026
Diffusion LLM at 1M context and 1000+ tok/s for real-time agents and long-horizon coding.
About this release
Live drop
InceptionDropped Oct 19, 2026
Diffusion LLM at 1M context and 1000+ tok/s for real-time agents and long-horizon coding.
About this release
Mercury 3 is Inception's flagship diffusion-based language model. It generates and refines tokens in parallel rather than autoregressively, delivering over 1,000 tokens per second on NVIDIA GPUs with a 1M-token context window. The model matches or exceeds quality of speed-optimized frontier models such as GPT-6 Luna and Claude Haiku 5.5 on selected benchmarks while remaining OpenAI-compatible via the Inception API.
Read the official Inception launch pageFirst builds incoming — this lane fills in automatically as people start building with Mercury 3. Know one? Submit the URL.