Inception launches Mercury 3, a diffusion LLM above 1000 tokens per second
The company says Mercury 3 outperforms GPT-6 Luna and Claude Haiku 5.5 on its own benchmarks, though Claude Haiku 5.5 leads on two others.
Inception released Mercury 3, a diffusion large language model that generates more than 1000 tokens per second, on October 19, 2026, according to aitimes.com. The company also introduced Mercury Voice and Mercury Decide alongside the model.
Measured generation speed on Nvidia GPUs was 1068 tokens per second. Mercury 3 supports a context length of up to 1 million tokens and has over 40 times more parameters than Mercury 1 Mini, Inception's previous-generation model.
Inception says Mercury 3 scores higher than OpenAI's GPT-6 Luna and Anthropic's Claude Haiku 5.5 on its own benchmarks, including 51.08 on SWE-bench Atlas versus 33.33 and 32.26 respectively, and 76.84 on IFBench, about 15 points above GPT-6 Luna. It scored 89.23 on GPQA Diamond. Claude Haiku 5.5 beat Mercury 3 on DeepSWE v1 (51.62 versus 46.31) and on MRCR v2.
Early partner results and availability
Early access partners reported their own figures. TinyFish said Mercury 3 completed 50 percent of 50 real web browsing tasks. Octen AI reported a median generation time of 0.9 seconds and a total response time of 2.2 seconds for smart-glasses-based search, while RevisionDojo said study hints and answer displays arrive in about 1 second.
Mercury 3 is available through the Inception API platform, OpenRouter, Baseten and Vercel. Standard pricing is $0.20 per million input tokens and $0.70 per million output tokens, with a 50 percent launch discount.
Mercury 1 Mini was Inception's first model, and average benchmark performance rose about 3x between it and Mercury 3.
Quick answers
How fast is Mercury 3?
Inception measured 1068 tokens per second on Nvidia GPUs, and the model generates more than 1000 tokens per second.
How much does Mercury 3 cost?
Standard pricing is $0.20 per million input tokens and $0.70 per million output tokens, with a 50 percent launch discount.
Where can I access Mercury 3?
It is available through the Inception API platform, OpenRouter, Baseten and Vercel.