Inception: Mercury 2.5

Inception: Mercury 2.5 is a text model from Inception. It belongs to the text category and is described by its provider as the latest diffusion LLM (dLLM) from the company, as well as the fastest reasoning LLM. The provider release date is 2026-09-08. Rather than generating tokens sequentially, Mercury 2.5 produces and refines multiple tokens in parallel.

The model accepts text as input and returns text as output. It has a context window of 260,000 tokens, which allows long documents, large code files or extended conversation histories to be handled in a single request. The maximum output is 65,536 tokens, so responses can be long when the task requires it. Because it is a reasoning model, it can work through problems before producing its final answer, and reasoning behaviour can be influenced through the supported parameters.

These capabilities point to a few practical uses. Long context work such as summarising or querying large documents, codebases or transcripts fits the 260,000 token window. Reasoning-heavy tasks, including multi-step analysis, planning and problem solving, are a natural match given the model's positioning as a reasoning LLM. The tool calling and structured output support make it suitable for agent style workflows and for applications that need responses in a defined format, such as extracting fields from text or driving downstream automation. Parallel token generation also makes it relevant where response latency matters.

You call the model through api14's OpenAI-compatible API using the model id "inception/mercury-2.5". The request parameters it supports are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, temperature, tool_choice and tools. In practice this means you can cap output length with max_tokens, control sampling with temperature, stop generation on given sequences with stop, request structured or formatted output through response_format and structured_outputs, and expose functions to the model with tools and tool_choice. The reasoning, reasoning_effort and include_reasoning parameters let you tune how much reasoning the model performs and whether reasoning content is returned.

There are cases where another model may fit better. Mercury 2.5 takes only text as input, so tasks involving images, audio or other modalities are outside its scope. Its output is also text only. If your workload does not benefit from reasoning or parallel generation, a simpler model may be sufficient, but within text tasks the context window, output limit and parameter set make Mercury 2.5 a flexible choice.