Inception

Inception Labs builds diffusion-based large language models that generate tokens in parallel instead of one at a time , and Mercury 2.5 is the company's latest production release for teams building latency-sensitive AI. It runs at 1,107 tokens per second on widely available NVIDIA GPUs and delivers a 40 percent jump in intelligence over Mercury 2, comparable to cost-optimized frontier models . The model ships with a 260K token context window, tunable reasoning, parallel tool calls, and schema-aligned JSON, aimed squarely at voice agents, search and RAG pipelines, and coding subagents where per-call latency compounds across a workflow. The launch matters because diffusion LLMs have moved from research curiosity to something real customers are running in production. Since Mercury 2 shipped, thousands of developers have built on it, dozens of enterprises have put it into production, and usage has grown by more than an order of magnitude . Inception cites concrete production wins to justify the release, including OpenCall bringing median model response latency close to 170 milliseconds for live phone agents, and Augment Code cutting context compaction latency by roughly 82 percent and cost by 90 percent when routing that work to Mercury. Inception is led by CEO Stefano Ermon, who announced the release. The founding team pairs Stanford professor Stefano Ermon, a co-inventor of diffusion models, with UCLA professor Aditya Grover and Cornell professor Volodymyr Kuleshov , and the company recently closed a $50M financing round led by Menlo Ventures, with participation from Andrew Ng and Andrej Karpathy among others . Mercury 2.5 is available through Inception's API, with launch pricing discounted 80 percent to $0.04 per million input tokens and $0.15 per million output.