News for hackers. Stuff that matters.

A comment on Mercury: Ultra-Fast Language Models Based on Diffusion.

Re:Mercury: Ultra-Fast Language Models Based on Diffusion by fastball on Monday July 07, 2025 @09:50AM (#44490372)
ICYMI, DeepMind also has a Gemini model that is diffusion-based[1]. I've tested it a bit and while (like with this model) the speed is indeed impressive, the quality of responses was much worse than other Gemini models in my testing.

[1] https://deepmind.google/models/gemini-diffusion/

[ Parent | on HN ]
Re:Mercury: Ultra-Fast Language Models Based on Diffusion by tripplyons on Monday July 07, 2025 @01:23PM (#44492577)
Is the Gemini Diffusion demo free? I've been on the waitlist for it for a few weeks now.
[ Parent | on HN ]
Re:Mercury: Ultra-Fast Language Models Based on Diffusion by rany_ on Monday July 07, 2025 @08:21PM (#44495836)
Yes it is.
[ Parent | on HN ]
Re:Mercury: Ultra-Fast Language Models Based on Diffusion by MattSayar on Tuesday July 08, 2025 @01:19PM (#44502043)
It took me a few weeks to get access. In the meantime Simon has a decent demo here https://simonwillison.net/2025/May/21/gemini-diffusion/
[ Parent | on HN ]
Re:Mercury: Ultra-Fast Language Models Based on Diffusion by Powdering7082 on Monday July 07, 2025 @01:26PM (#44492612)
From my minor testing I agree that it's crazy fast and not that good at being correct
[ Parent | on HN ]