Inception Debuts Mercury 2.5 Diffusion Model With $50M Funding

Inception has released Mercury 2.5, a diffusion-based large language model that improves intelligence by 40% compared to the previous Mercury 2 version. The model processes more than 1,100 tokens per second and utilizes a non-autoregressive approach, generating multiple tokens simultaneously rather than sequentially.

The model features a 260K token context window, capable of processing the equivalent of a 500-page book in one prompt. It includes tunable reasoning levels for adjusting thinking depth, parallel tool calls, and structured JSON output. Inception positions Mercury 2.5 as a competitor to GPT-5.6 Luna (Low) and Gemini 3.5 Flash-Lite.

Standard pricing is set at $0.20 per million input tokens and $0.75 per million output tokens. During the launch phase, an 80% discount reduces these costs to $0.04 per million input tokens and $0.15 per million output tokens. Integration is simplified via an OpenAI-compatible API.

The startup has raised $50 million in funding led by Menlo Ventures, with participation from Andrew Ng and Andrej Karpathy. Early implementation by voice AI company OpenCall showed median latencies under 200 milliseconds for voice interactions. The model is designed for voice agents, coding assistants, and enterprise search.