Inception: Mercury 2

Inception: Mercury 2 is a text model from Inception, listed on api14 under the model id inception/mercury-2. It belongs to the text category and is described by its provider as an extremely fast reasoning LLM, and as the first reasoning diffusion LLM (dLLM). The provider release date is 2026-03-04. If you are choosing a model for a text pipeline, the distinguishing detail here is the diffusion approach to generation rather than a conventional sequential decoder.

On capabilities, Mercury 2 takes text in and returns text out, so it is not a multimodal model. It has a context window of 128,000 tokens and a maximum output of 50,000 tokens. The provider description notes that instead of generating tokens sequentially, Mercury 2 produces and refines multiple tokens in parallel. Reasoning behaviour is exposed through request parameters rather than being fixed, which matters if you want to control how much internal reasoning the model performs before it answers.

The practical use cases follow from that combination of a large context window, a large output budget and configurable reasoning. Long documents, multi file codebases, transcripts and research material fit inside 128,000 tokens, and the 50,000 token output ceiling supports long form answers, structured reports and extended reasoning traces. Because the model accepts tools and structured outputs, it also suits agent style workflows and applications that need machine readable responses rather than free text. Teams that care about latency in reasoning heavy tasks are the natural audience for a diffusion based reasoning model.

You call it through api14's OpenAI-compatible API using the model id inception/mercury-2. The supported request parameters are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, temperature, tool_choice and tools. In practice that means you can set a token ceiling with max_tokens, tune sampling with temperature, constrain generation with stop sequences, request JSON or schema shaped output through response_format and structured_outputs, and attach function calling through tools and tool_choice. The reasoning, reasoning_effort and include_reasoning parameters let you decide how much reasoning the model does and whether that reasoning is returned to you.

The facts given do not list unsupported modalities, rate limits or pricing behaviour, so treat the model as text only and check the live price table on the page for cost. If your task needs image or audio input, Mercury 2 is not the right fit, since its input and output modalities are text alone. For straightforward text generation where you do not need reasoning control, tools or structured output, a simpler model may be sufficient, but within the stated facts Mercury 2 covers long context text work, tool use and schema constrained responses.