Sam Witteveen · 2026-10-20 · notable
Sam Witteveen — 'Do Diffusion LLMs Scale? Testing Mercury 3 + Decide'
Sam Witteveen tests Inception's diffusion LLM Mercury 3 and the Mercury Decide decision model, with speed runs, a 5,000-word essay, tool calling, RAG, image-based decisions and launch pricing. Inception sponsored the video.
Sam Witteveen puts Inception's Mercury 3 diffusion LLM and Mercury Decide through speed, agent and RAG tests.
What is it?
'Do Diffusion LLMs Scale? Testing Mercury 3 + Decide' went up on Sam Witteveen's channel on 20 October 2026. The video looks at Mercury 3, Inception's diffusion-based LLM, and Mercury Decide, a decision model that can now read images. Inception sponsored it.
How does it work?
Chapters explain how diffusion LLMs generate text, then cover benchmarks, swarms of sub-agents, speed, OpenAI compatibility and the 1M context. The hands-on half runs a real-time speed test, a 5,000-word essay, agentic tool-calling, a RAG test, image-based decisions and a pipeline that chains Mercury Decide into Mercury 3, ending with launch pricing.
Why does it matter?
Diffusion LLMs promise far higher speed than token-by-token models, and the open question is whether they hold up on real agent work. A run through tool calling and RAG shows builders where Mercury 3 fits before they switch a pipeline to it.
Who is it for?
developers evaluating fast LLMs for agents