Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Mercury: Commercial-scale diffusion language model

inceptionlabs.ai

Post reply on HN
Print view View original

31–40 of 189 posts

Re: Mercury: Commercial-scale diffusion language model

#31

Post by kuhewa »

whall6 wrote: post #9
marcyb5st wrote: post #6

Super happy to see something like this getting traction. As someone that is trying to reduce my carbon footprint sometimes I feel bad about asking any model to do something trivial. With something like that perhaps the guilt will lessen

If you live in the U.S., marginal electricity demand during the day is almost invariably met with solar or wind (solar typically runs at a huge surplus on sunny days). Go forth and AI in peace, marcyb5st.

This made me wonder - do any cloud compute systems have an option to time jobs or use physical resources geographically based on surplus power availability to minimise emissions?

I reckon it might incidentally happen if optimising for cost of power depending how correlated that is to carbon intensivity of power generation, which admittedly I haven't thought through.

Re: Mercury: Commercial-scale diffusion language model

#32

Post by drusepth »

jonplackett wrote: post #5

Ok. My go to puzzle is this: You have 2 minutes to cool down a cup of coffee to the lowest temp you can You have two options: 1. Add cold milk immediately, then let it sit for 2 mins. 2. Let it sit for 2 mins, then add the cold milk. Which one cools the coffee to the lowest temperature and why? And Mercury gets this right - while as of right now ChatGPT 4o get it wrong. So that’s pretty impressive.

Is there a trick question in here, where the cold milk is trending closer to room temperature over the two minutes waited?

Re: Mercury: Commercial-scale diffusion language model

#33

Post by cratermoon »

jonplackett wrote: post #5

Ok. My go to puzzle is this: You have 2 minutes to cool down a cup of coffee to the lowest temp you can You have two options: 1. Add cold milk immediately, then let it sit for 2 mins. 2. Let it sit for 2 mins, then add the cold milk. Which one cools the coffee to the lowest temperature and why? And Mercury gets this right - while as of right now ChatGPT 4o get it wrong. So that’s pretty impressive.

> My go to puzzle is this:

> Mercury gets this right - while as of right now ChatGPT 4o get it wrong.

This is so common a puzzle it's discussed all over the internet. It's in the data used to build the models. What's so impressive about a machine that can spit out something easily found with a quick web search?

Re: Mercury: Commercial-scale diffusion language model

#34

Post by byearthithatius »

nullc wrote: post #15
byearthithatius wrote: post #4

Interesting approach. However, I never thought of auto regression being _the_ current issue with language modeling. If anything it seems the community was generally surprised just how far next "token" prediction took us. Remember back when we did char generating RNNs and were impressed they could make almost coherent sentences? Diffusion is an alternative but I am having a hard time understanding the whole "built in…

Consider the entropy of the distribution of token X in these examples: "Four X" and "Four X and seven years ago". In the first case X could be pretty much anything, but in the second case we both know the only likely completion. So it seems like there would be a huge advantage in not having to run autogressively. But in practice it's less significant then you might imagine because the AR model can internally model th…

Both are conditional distributions on the context of which they were requested so like you said in the second paragraph, the difference is not significant. I see what you mean though and maybe there are use cases then where Diffusion is preferable. To me it seems the context conditional and internal model is sufficient where this problem doesn't really occur.

Re: Mercury: Commercial-scale diffusion language model

#35

Post by AlexCoventry »

XenophileJKO wrote: post #29
inerte wrote: post #7

Not sure if I would tradeoff speed for accuracy. Yes, it's incredible boring to wait for the AI Agents in IDEs to finish their job. I get distracted and open YouTube. Once I gave a prompt so big and complex to Cline it spent 2 straight hours writing code. But after these 2 hours I spent 16 more tweaking and fixing all the stuff that wasn't working. I now realize I should have done things incrementally even when I hav…

So my personal belief is that diffusion models will enable higher degrees of accuracy. This is because unlike an auto-regressive model it can adjust a whole block of tokens when it encounters some kind of disjunction. Think of the old example where an auto regressive model would output: "There are 2 possibilities.." before it really enumerated them. Often the model has trouble overcoming the bias and will hallucinate…

> it can adjust a whole block of tokens when it encounters some kind of disjunction.

This is true in principle for general diffusion models, but I don't think it's true for the noise model they use in Mercury (at least, going by a couple of academic papers authored by the Inception co-founders.) Their model generates noise by masking a token, and once it's masked, it stays masked. So the reverse-diffusion gets to decide on the contents of a masked token once, and after that it's fixed.

Re: Mercury: Commercial-scale diffusion language model

#36

Post by behnamoh »

jonplackett wrote: post #5

Ok. My go to puzzle is this: You have 2 minutes to cool down a cup of coffee to the lowest temp you can You have two options: 1. Add cold milk immediately, then let it sit for 2 mins. 2. Let it sit for 2 mins, then add the cold milk. Which one cools the coffee to the lowest temperature and why? And Mercury gets this right - while as of right now ChatGPT 4o get it wrong. So that’s pretty impressive.

Qwen3 30B A3B: https://pastebin.com/kdJZvqVb

Re: Mercury: Commercial-scale diffusion language model

#37

Post by dmos62 »

If the benchmarks aren't lying, Mercury Coder Small is as smart as 4o mini and costs the same, but is order of magnitude faster when outputting (unclear if pre-output delay is notably different). Pretty cool. However, I'm under the impression that 4o-mini was superceded by 4.1-mini and 4.1-nano for all use cases (correct me if I'm wrong). Unfortunately they didn't publish comparisons with the 4.1 line, which feels like an attempt to manipulate the optics. Or am I misreading this?

Btw, why call it "coder"? 4o-mini level of intelligence is for extracting structured data and basic summaries, definitely not for coding.

Re: Mercury: Commercial-scale diffusion language model

#38

Post by macleginn »

AlexCoventry wrote: post #35
XenophileJKO wrote: post #29

Earlier quoted context omitted.

So my personal belief is that diffusion models will enable higher degrees of accuracy. This is because unlike an auto-regressive model it can adjust a whole block of tokens when it encounters some kind of disjunction. Think of the old example where an auto regressive model would output: "There are 2 possibilities.." before it really enumerated them. Often the model has trouble overcoming the bias and will hallucinate…

> it can adjust a whole block of tokens when it encounters some kind of disjunction. This is true in principle for general diffusion models, but I don't think it's true for the noise model they use in Mercury (at least, going by a couple of academic papers authored by the Inception co-founders.) Their model generates noise by masking a token, and once it's masked, it stays masked. So the reverse-diffusion gets to dec…

The exact types of path dependencies in inference on text-diffusion models look like an interesting research project.

Re: Mercury: Commercial-scale diffusion language model

#39

Post by XenophileJKO »

AlexCoventry wrote: post #35
XenophileJKO wrote: post #29

Earlier quoted context omitted.

So my personal belief is that diffusion models will enable higher degrees of accuracy. This is because unlike an auto-regressive model it can adjust a whole block of tokens when it encounters some kind of disjunction. Think of the old example where an auto regressive model would output: "There are 2 possibilities.." before it really enumerated them. Often the model has trouble overcoming the bias and will hallucinate…

> it can adjust a whole block of tokens when it encounters some kind of disjunction. This is true in principle for general diffusion models, but I don't think it's true for the noise model they use in Mercury (at least, going by a couple of academic papers authored by the Inception co-founders.) Their model generates noise by masking a token, and once it's masked, it stays masked. So the reverse-diffusion gets to dec…

Thank you, I'll have to read the papers. I don't think I have read theirs.

Re: Mercury: Commercial-scale diffusion language model

#40

Post by jtonz »

I would be interested to see how people would apply this working as a coding assistant. For me, its application in solutioning seem very strong, particularly vibe coding, and potentially agentic coding. One of my main gripes with LLM-assisted coding is that for me to get the output which catches all scenarios I envision takes multiple attempts in refining my prompt requiring regeneration of the output. Iterations are slow and often painful.

With the speed this can generate its solutions, you could have it loop through attempting the solution, feeding itself the output (including any errors found), and going again until it builds the "correct" solution.

Post reply on HN