LanguageSearch

Mercury

#9

inception-labs · seit Februar 2025 · 5× · zuletzt 29. Juni 2026

28

Momentum

Mercury is a family of commercial-scale diffusion-based large language models from Inception Labs that generates text through parallel iterative refinement rather than sequential token prediction. The model achieves over 1,000 tokens per second on NVIDIA H100 GPUs and is up to 10 times faster than comparable autoregressive models while maintaining comparable quality. Mercury is provided via API, on-premise deployments, and a web interface.

Preuves (5)