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

Grok 4 Fast now has 2M context window

docs.x.ai

Post reply on HN
Print view View original

221–230 of 328 posts

Re: Grok 4 Fast now has 2M context window

#221

Post by Hamuko »

cbm-vic-20 wrote: post #220
Hamuko wrote: post #211

Earlier quoted context omitted.

Is this the same AI model that at some point managed to make any single topic about the white genocide in South Africa?

How does this sort of thing work from a technical perspective? Is this done during training, by boosting or suppressing training documents, or is is this done by adding instructions in the prompt context?

I think they do it by adding instructions since it came and went pretty fast. Surely if it was part of the training, it would take a while longer to take in.

Re: Grok 4 Fast now has 2M context window

#222

Post by Havoc »

jorvi wrote: post #193
mehdibl wrote: post #8

What matter is not context or the recod token/s you get. But the quality for the model. And it seem Grok pushing the wrong metrics again, after launching fast.

Grok's biggest feature is that unlike all the other premier models (yes I know about ChatGPT's new adult mode), it hasn't been lobotomized by censoring.

No censoring and it says the things I agree with are not the same thing

Re: Grok 4 Fast now has 2M context window

#223

Post by TheDong »

basisword wrote: post #198
jorvi wrote: post #193

Earlier quoted context omitted.

Grok's biggest feature is that unlike all the other premier models (yes I know about ChatGPT's new adult mode), it hasn't been lobotomized by censoring.

I’ve never run into this problem. What are you asking LLM’s where you run it censoring you?

I sometimes use LLM models to translate text snippets from fictional stories from one language to another.

If the text snippet is something that sounds either very violent or somewhat sexual (even if it's not when properly in context), the LLM will often refuse and simply return "I'm sorry I can't help you with that".

Re: Grok 4 Fast now has 2M context window

#224

Post by benzible »

cbm-vic-20 wrote: post #220
Hamuko wrote: post #211

Earlier quoted context omitted.

Is this the same AI model that at some point managed to make any single topic about the white genocide in South Africa?

How does this sort of thing work from a technical perspective? Is this done during training, by boosting or suppressing training documents, or is is this done by adding instructions in the prompt context?

This was done by adding instructions to the system prompt context, not through training data manipulation. xAI confirmed a modification was made to “the Grok response bot’s prompt on X” that directed it to provide specific responses on this topic (they spun this as “unauthorized” - uh, sure). Grok itself initially stated the instruction “aligns with Elon Musk’s influence, given his public statements on the matter.” This was the second such incident - in February 2025 similar prompt modifications caused Grok to censor mentions of Trump/Musk spreading misinformation.

[1] https://techcrunch.com/2025/05/15/xai-blames-groks-obsession...

Re: Grok 4 Fast now has 2M context window

#225

Post by throwuxiytayq »

d4rkp4ttern wrote: post #208
changoplatanero wrote: post #2

Anyone can make a long context window. The key is if your model can make effective use of it or not.

There are “needle in the haystack” benchmarks for long context performance. It would be good to see those.

These aren’t really indicative of real world performance. Retrieving a single fact is pretty much the simplest possible task for a long context model. Real world use cases require considering many facts at the same time while ignoring others, all the while avoiding the overall performance degradation that current models seem susceptible to when the context is sufficiently full.

Re: Grok 4 Fast now has 2M context window

#226

Post by throwuxiytayq »

mg wrote: post #175
MallocVoidstar wrote: post #160

Earlier quoted context omitted.

I don't think there are any up-to-date leaderboards, but models absolutely degrade in performance the more context they're dealing with. https://wandb.ai/byyoung3/ruler_eval/reports/How-to-evaluate... >Gpt-5-mini records 0.87 overall judge accuracy at 4k [context] and falls to 0.59 at 128k. And Llama 4 Scout claimed a 10 million token context window but in practice its performance on query tasks drops below 20% accur…

That makes me wonder if we could simply test this by letting the LLM add or multiply a long list of numbers? Here is an experiment: https://www.gnod.com/search/#q=%23%20Calcuate%20the%20below%... The correct answer: Correct: 20,192,642.460942328 Here is what I got from different models on the first try: ChatGPT: 20,384,918.24 Perplexity: 20,000,000 Google: 25,167,098.4 Mistral: 200,000,000 Grok: Timed out after 300s…

I’m starting to find it unreasonably funny how people always want language models to multiply numbers for some reason. Every god damn time. In every single HN thread. I think my sanity might be giving out.

Re: Grok 4 Fast now has 2M context window

#227

Post by neidu »

basisword wrote: post #198
jorvi wrote: post #193

Earlier quoted context omitted.

Grok's biggest feature is that unlike all the other premier models (yes I know about ChatGPT's new adult mode), it hasn't been lobotomized by censoring.

I’ve never run into this problem. What are you asking LLM’s where you run it censoring you?

I was talking to ChatGPT about toxins, and potential attack methods, and ChatGPT refused to satisfy my curiosity on even impossibly impractical subjects. Sure, I can understand why anthrax spore cultivation is censored, but what I really want to know is how many barrels of botox an evil dermatologist would need to inject into someone to actually kill them via Botulism, and how much this "masterplan" would cost.

Re: Grok 4 Fast now has 2M context window

#228

Post by felixgallo »

neidu wrote: post #227
basisword wrote: post #198

Earlier quoted context omitted.

I’ve never run into this problem. What are you asking LLM’s where you run it censoring you?

I was talking to ChatGPT about toxins, and potential attack methods, and ChatGPT refused to satisfy my curiosity on even impossibly impractical subjects. Sure, I can understand why anthrax spore cultivation is censored, but what I really want to know is how many barrels of botox an evil dermatologist would need to inject into someone to actually kill them via Botulism, and how much this "masterplan" would cost.

[flagged]

Re: Grok 4 Fast now has 2M context window

#229

Post by sd9 »

jorvi wrote: post #193
mehdibl wrote: post #8

What matter is not context or the recod token/s you get. But the quality for the model. And it seem Grok pushing the wrong metrics again, after launching fast.

Grok's biggest feature is that unlike all the other premier models (yes I know about ChatGPT's new adult mode), it hasn't been lobotomized by censoring.

I am amazed people actually believe this

Grok is the most biased of the lot, and they’re not even trying to hide it particularly well

Re: Grok 4 Fast now has 2M context window

#230

Post by simondotau »

wqaatwt wrote: post #74
RobKohr wrote: post #32

Earlier quoted context omitted.

I like grok for noncoding stuff. I find it hasn't been tuned for "Safety" (meaning it isn't tuned much for political correctness). It also seems good at making images and stories up well. I run some choose your own adventures stories with my kids through it. We tell it who each of their characters are and what the theme is for the night and grok gives them each a section of story and 4 choices. They also have the opt…

> it isn't tuned much for political correctness It was tuned to be edgy and annoying though (I mean his general style of speech not necessarily the content).

Nothing in AI is more edgy and annoying than beginning every response with a mandatory glazing, like ChatGPT. “That’s a really insightful question, and shows that you really understand the subject!”
Post reply on HN