Inception released Mercury 3 on October 19, 2026, calling it the most capable diffusion-based large language model released to date. Why it matters: Mercury 3 runs at more than 1,000 tokens per second, offers a one-million-token context window, and launches at $0.10 per million input tokens and $0.35 per million output tokens — a speed-and-price combination aimed squarely at real-time agents.
October 19, 2026
What happened
Inception introduced Mercury 3 as its flagship diffusion LLM. Unlike standard autoregressive models, which generate text one token at a time in sequence, diffusion models generate text in a different way that can be much faster. Inception says Mercury 3 outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks while running four to eight times faster, at 1,068 tokens per second on NVIDIA GPUs. Those benchmark claims are company-reported and have not been independently verified.
The context window is one million tokens, up from 260,000 tokens for Mercury 2.5. List pricing is $0.20 per million input tokens and $0.70 per million output tokens, with a 50% launch discount bringing the effective rates to $0.10 and $0.35. Cached input tokens are priced 90% below the input rate.
Does diffusion scale?
Inception framed the release as evidence that text diffusion scales. The company says Mercury 3 is more than 40 times larger than Mercury 1 in total parameters, while measured intelligence across standard benchmarks rose threefold from Mercury 1 (mini) to Mercury 3. The open question, in Inception’s telling, was whether diffusion LLMs could scale to compete with autoregressive models on complex work — and the company says Mercury 3 answers it.
Early use in production
Inception says early-access partners are already running Mercury 3 in latency-sensitive work including browser-use agents, interactive education, real-time search agents, and customer support. In its own production web-agent harness, the company says Mercury 3 completed 50% of 50 live Mind2Web navigation tasks — a count it described as significantly higher than GPT-6.1-Sol’s, at the same token volume and number of agent steps and a fraction of the per-token cost. Partner comments in the announcement highlighted instant student feedback in tutoring and sub-second responses for smart-glasses use.
Mercury 3 is available through the Inception API Platform and through OpenRouter, Baseten, and Vercel, and is OpenAI API compatible. New platform accounts include 100 million free tokens. Mercury 2 and Mercury 2.5 are no longer available to new users; existing customers keep access until November 15, 2026.
Why it matters
Most AI products people actually wait on — voice agents, tutoring, live search, support chats — are limited by latency, not just intelligence. A model that is genuinely fast at over 1,000 tokens per second, cheap at launch pricing, and competitive on benchmarks would change what those products can do in real time. The honest test is the same as always: Inception’s benchmark and speed claims are its own, so run Mercury 3 on your own latency-sensitive tasks before switching anything that matters.
Frequently asked questions
What is Mercury 3?
Mercury 3 is Inception’s flagship diffusion-based large language model, released October 19, 2026, built for real-time, latency-sensitive applications.
How fast is Mercury 3 and what does it cost?
Inception lists 1,068 tokens per second on NVIDIA GPUs. Launch pricing is $0.10 per million input tokens and $0.35 per million output tokens, a 50% discount on list prices of $0.20 and $0.70.
What is Mercury 3’s context window?
One million tokens, up from 260,000 tokens for Mercury 2.5.
Where can you use Mercury 3?
Through the Inception API Platform, OpenRouter, Baseten, and Vercel. It is OpenAI API compatible, and new Inception accounts include 100 million free tokens.
Sources: Unite.AI reporting on Inception’s announcement (October 19, 2026)