Hacker News

points by RajuChacha108 1 day ago | 0 comments

It is pretty clear at this point that the frontier models need only handful of tiered models.

1. Ultra - The best and the most expensive for showing off and government. 2. Pro - Almost best but substantially cheaper for most productive people. 3. Flash - Workhorse for everyone. 4. Lite - For the unwashed masses and peasants.

There is no need to attach anything to these labels. In fact industry could simply standardize these namings like SEMVER.

There could be an internal representation which could give specific number of the model but these companies should move on to be more like Chrome. No one knows or cares which version of chrome they are on.

For a lot of systems the real underlying version would matter because of reliability. That can be provided to those power users.

[–] blauditore 1 day ago

You could say the same about many products, but marketing is often about not making the tiering obvious. Just try to make sense of the naming strategy of Samsung TVs, and you'll either eventually give up or go insane.

[–] Dagonfly 23 hours ago

TVs and monitors have this naming strategy because of retail stores. You can't price compare between different stores if every one has their own exclusive model with a different set of RNG numbers in the name.

Same for appliances or most products that are bought through a middleman rather than directly from the manufacturer.

[–] simondotau 1 day ago

It makes humans go insane. On the other hand it’s perfect for AI research.

[–] mrheosuper 1 day ago

Is it as bad as PC monitor naming?

[–] epolanski 1 day ago

Very complex naming is a feature in the entire monitor industry.

[–] Uhhrrr 1 day ago

There's a tier which is just, "summarize things and do grammar checks, style tweaks, and boilerplate generation". Maybe this is Lite. It's like how computing requirements for most people nowadays is "run a browser".

[–] notatoad 1 day ago

>There's a tier which is just, "summarize things and do grammar checks, style tweaks, and boilerplate generation"

honestly, this tier is the 4GB model that ships in chrome and runs locally

[–] mrits 1 day ago

It's a bit sad to think that 100 years from now the low tier models will still be better than us at grammar. They are getting better and we aren't

[–] Marsymars 1 day ago

> It's a bit sad to think that 100 years from now the low tier models will still be better than us at grammar.

Not if you're a linguistic descriptivist.

[–] aerodexis 1 day ago

depends on what being good at grammar entails. If it's just following rules, then you're not wrong. If it involves adapting concepts to novel circumstances, we're doing well.

[–] Obscurity4340 12 hours ago

I aint sad 'bout nuthin

[–] paradox460 1 day ago

Not even for government. Many of these models aren't available on fedramp available providers, like bedrock. Instead they're stuck months ago. Iirc bedrock just got sonnet 5

[–] Razengan 1 day ago

Excuse me, I am a washed peasant thank you

[–] switchbak 1 day ago

You're in-between tiers, and thus don't exist. Please wait 4 days between washing and try again.

[–] Duwensatzaj 1 day ago

Pro v Flash - I always thought of Flash as quick, Pro as slower, Chain of Thought enabled.

[–] smallerize 1 day ago

But Gemini (the app/website) lets you set the Thinking level for Flash or even Flash-Lite. And Flash with Low thinking is the default.

[–] 9dev 1 day ago

And that’s exactly where it starts to get murky, because this isn’t a clear gradient after all but a matrix of different qualities for different tasks. Reasoning effort levels are one factor here, but the whole speed vs. Quality topic is another.

[–] emayljames 23 hours ago

From using Gemini, flash needs a lot of corrective input, pro rarely does

[–] yipinwong 1 day ago

I do not agree.

As an example, Linux distros say, Ubuntu got its project "names" as well as SEMVERs. But people rarely use project names but uses versions.

You tell me, I use "Jammy Jellyfish". I will be like WTF dood. You tell me 22.04 LT (which is one I use), oh yeah, that's a good one.

Same thing here. companies can call whatever SOTA whatever they want. But in the end people will refer to them as just Gemini 4 or Ultr/Pro/Lite however people like to use.

Just leave it to company's discretion. Don't enforce anything. People will figure out how to call it.

---

And wtf, a peaseant?

I use your so-stupidly-called ultra/med/lite but in diff situations, and when I use lite that doesn't mean I am peasant, I am being sensible

[–] smellf 1 day ago

Ubuntu's release numbers are not semver, they're year and month. 22.04 is the April release from 2022. Your LT is redundant, all 04 releases from even-numbered years are LTS. And people do frequently refer to Ubuntu releases by the code name, though usually just the adjective and not the animal - jammy, noble, etc.

[–] redox99 1 day ago

For debian it's quite common to use the toy story name.

[–] gorgolo 1 day ago

I think he was referring to us pheasants.