Inception Chatv2

Adds an explicit “reasoning” positioning, described as the fastest reasoning language model in their lineup. Introduces tunable reasoning as a first-class feature for trading speed vs depth per request. Adds native tool use plus schema-aligned JSON output as headline production features. Sharpens the “production loops” focus (agents, RAG, extraction) as the core value proposition rather than just raw fast generation. OpenAI-API compatibility so it can slot into existing stacks without rewrites.

First commercial-scale diffusion LLM family designed for fast, high-quality text generation. “Mercury Coder” is a code-focused variant intended for coding benchmarks, RAG, tool use, and agent workflows. Parallel, coarse-to-fine generation (diffusion) instead of token-by-token decoding, enabling much faster throughput. Positioned as a drop-in replacement for autoregressive LLM use cases, including tool use and agentic workflows. Emphasis on speeding up long reasoning traces and making “reasoning-grade” behavior more practical at lower latency

Initial release of Inception Chat.