News for hackers. Stuff that matters.
Re:Mercury 2: The fastest reasoning LLM, powered by diffusion by micw on Wednesday February 25, 2026 @12:35AM (#47147727)
On a very old model, it's more like 16.000 garbage words/s
Re:Mercury 2: The fastest reasoning LLM, powered by diffusion by nl on Wednesday February 25, 2026 @12:58AM (#47147874)
Llama 3.1 8B is pretty useful for some thing. I use it to generate SQL pretty reliably for example.

They are doing an updated model in a month or so anyway, then a frontier level one "by summer".

Re:Mercury 2: The fastest reasoning LLM, powered by diffusion by numeri on Wednesday February 25, 2026 @10:24AM (#47152765)
but Taalas had to quantize Llama 3.1 8B to death to get it to fit. It can't produce coherent non-English text at all.
Re:Mercury 2: The fastest reasoning LLM, powered by diffusion by patapong on Wednesday February 25, 2026 @06:01AM (#47149970)
I do wonder if there are tasks where 16k garbage words/s are more useful than 200 good words per second. Does anyone have any ideas? Data extraction perhaps?
Re:Mercury 2: The fastest reasoning LLM, powered by diffusion by pnocera on Wednesday February 25, 2026 @02:45PM (#47156782)
A politician communication agent maybe...