Inception Labs
The company behind the Mercury family of diffusion LLMs, sold through the Inception Platform API (inception-platform-docs). It is this spoke's first provider selling a non-autoregressive text model commercially.
What is known here comes from the product documentation only: three models (Mercury 2.5, Mercury 2, Mercury Edit 2), an OpenAI-compatible API, a promotional price, and a latency-first pitch aimed at voice and inline code completion. No funding, founders, headcount or history are recorded — the docs name none, and nothing else was verified at ingest, so no Wikipedia or Wikidata identifier is claimed.
The strategic position is narrow and legible: not competing on capability rankings (llm-benchmarks holds no Mercury result), but on the two workloads where autoregressive decode is felt most directly by a user — a cursor waiting for a completion, and a voice agent waiting to speak.