Cursor, since Grok 4.5, has had an incredible deal for frontier level models, their subscription now goes way further than OpenAI or Anthropic. Even on their lower tier plans you can use a lot tokens on their of their first party models (Grok and Composer) and not really run out comparatively. Combine them with an orchestrator and implementor type setup and it goes even further.
Can you explain what you mean? These days courtesy of an addictive reset game OpenAI is playing, I can't find anything with frontier intelligence that's more cost efficient...
If they didn’t constantly reset, they’d be about the same as Anthropic.
Right now, I find that Grok offers better value, uses fewer tokens per turn, and makes better code. I haven’t tried Cursor because I don’t want to change editors again, but maybe I should try it…
Not that I know of. AA's token use metrics (mentioned in this article) are indicative, however. They say explicitly here that the Grok models are notably token efficient. This is my experience.
I believe they are the only western provider that has Kimi K3 on a subscription plan today as well. I would love to ditch Anthropic and be on Kimi if there were a subsidized plan like that with ZDR
I’d love a subsidized Kimi subscription too. The official Kimi subscription is always out of stock and doesn’t have great limits, while the K3 allotments on OpenCode and Cursor don’t seem to last very long either.
They don't do request based pricing anymore. Its just token based (1 credit = $0.01) plus some bonus credit based on which plan you subscribe. So for example a $39 plan get $70 of credits.
Yeah, I know, but "credit" translates differently because the models bill at different rates, which gets turned into "multipliers" (or at least, it did).
Have they converted entirely to transparent API rates + base allocation now? One of the reasons I left was that if I was going to be billed at API rates anyway, I'd just rather use the APIs. The value proposition still sucks for individuals now, when the other major providers are bundling at below-API rates.
I'm willing to pay 2x for a 10% smarter model. Intelligence matters that much (because 10% smarter probably saves, on average, several hours of human time).
For building a full stack custom CRM and media pipeline tool with video conversion, transcription, and indexing. Supabase, AWS, Meili, NextJS, GCS - lots of surfaces and planes.
4.8 basically couldn't do it, I abandoned the project as the fallback was, "current business processes".
With F5 it's been 4 weeks and almost ready for production release.
SpaceXAI is the only frontier model company that had its own compute/date centres and soon chip making factory, I think they will pull ahead with cheaper tokens similar intelligence and better harness/tools. Grok build is 2-5x faster than Claude Code in my opinion.
OpenAI is almost there, and Anthropic is pretty close behind. In the next year or two all major AI companies will be vertically integrated to a good degree.
> I think they will pull ahead with cheaper tokens similar intelligence
They obviously have a huge token cost advantage of the AI labs they are renting compute to, at least for now while they can charge current crazy rates for GPU compute.
Except EUV lithography is the most complicated industrial process that exists and they won't have usable yields for many years if ever. I don't think Musk actually expects these fans to ever actually make sense they just let him hype and distract.
Light generation is the most complicated part and Elon plans on doing Free Electron Laser (FEL) which is not as complicated as self contained tin based solution that ASML uses now
I recently decided to get an AI subscription and evaluated Grok vs chatgpt. Went with Grok because it's all-around good enough at day to day stuff, integrates with my Tesla, and the image/video generation is great. Kids love whimsical videos of them riding dinosaurs.
Feedback: I'd like Grok to have more connectors (I see OpenAI just added Apple Health, that would be nice to have, and I wish it could read my Onenote notebooks) and for existing ones to be improved. I gave it access to my gmail and asked it "what was my last electricity bill?". It failed to find it, even when I told it the exact subject line to search for. Something about not getting any data back when trying to get the email contents.
Grok doesn’t have those features, and people who like to make adult content have been complaining for a while how Grok has made it a lot harder to do so.
note that Grok's training, thanks to its portable gas generators that are magnitudes less efficient than even other integrated, permanent gas turbines, means the training for this model is dramatically less efficient than models like DeepSeek
a lot of the CO2 emission debate on AI is overblown but it's accurate for Grok
I have. He was using it due to philosophical reasons the same way many people have philosophical reasons for avoiding it. I don't know how many people are like that, but it's not exactly where you want to position your product if you're a business.
Personally - and I know I'm not alone with this sentiment based on comments I see on this site - I wouldn't touch Grok no matter how good or cheap it is. I don't trust Elon and I don't want to give another dollar to the world's richest person who turns around and uses the money to interfere with elections. The guy I know uses it for essentially the same reason I won't use it.
Perhaps customers choosing your product for irrational philosophical reasons is exactly how you'd want to position your product if you're a business. When it comes to margins, the only thing better than a price-insensitive customer is a quality-insensitive customer.
> the world's richest person who turns around and uses the money to interfere with elections.
I'm no Elon fanboy but to be fair a lot of the richest people in the world heavily fund and endorse political candidates, parties, PACs, lobbyists, etc across the political spectrum. Elon was just far more public about it and it was covered much more.
This statement doesn't go far enough given Elon's direct and hands-on involvement with DOGE and the 2024 elections. Very few of the richest people of the world are personally entangled in meddling with government agencies directly, for example.
Elon Musk has actively endorsed the right wing extremist party in Germany.
The one that often has trouble running it's events as businesses try to avoid serving them. And which has prominent members that the secret service considers definitely fascist.
Musk has gone a long way beyond simply being the standard rich person lobbying for their own interests, and has moved into actively promoting people that are trying to tear down liberal democracy.
> MADISON, Wis. (AP) — Billionaire Elon Musk likely broke Wisconsin law when he promised to hand out $1 million checks to voters in the 2025 state Supreme Court election, a bipartisan panel has found.
> The Wisconsin Elections Commission last week referred two complaints to the Brown County district attorney’s office, which can choose to bring criminal charges over violating the state law against election bribery. Prosecutors have 40 days to report back to the commission.
Journalism has gone wild these days. He's already been sued and won a couple times. An for what? Encouraging people just to vote (he did not say which way). It's going to backfire hard for all the ad spend by MTV, Meta, and Google who also spend lots of money to encourage certain demographics to vote.
Idk I just don't like the richest man on planet and the owner of the "town square of the internet" to post fake news blatantly to promote hate against a group of people.
Versus Sam,Dario or the CCP? Im all for running local models but im sor far from being able to pay for a large model hardware setup. My strix halo box is like driving a beaten up vespa when the frontier models are Ferraris.
I probably have the same strix halo box as you. It's slow, although mostly tolerable, but even 128 GB isn't enough to run good models. Where that leaves us is giving money to somebody to get access to frontier models. I don't like any of those guys either, but some appear worse than others. FWIW, I mostly use Anthropic models. And CCP and their distilled models notwithstanding, at least they release open weight models at a lower cost. It's not like any American company can claim the higher ground these days anyway, especially when they trained their models on pirated content and trash our neighborhoods with their datacenters.
Yes, ethically Elon ranks lowest based on the past 5 years of questionable decisions, like uploading your entire codebase to their server without consent.
Elon is directly responsible for Grok becoming self-titled “MechaHitler” which the other three haven’t come close to matching yet.
There's a big difference between quiet donations to a PAC - not that that's good either - and what Elon did. He literally paid for votes, likely in violation of the law. He poured more money into U.S. elections than anyone has before. He and Trump both made strange, cryptic statements about Elon's role in Pennsylvania with the voting machines that has caused people to reasonably wonder if they somehow manipulated the election. Whether he did or not, the innuendo alone is not ok. Then he did what he did in Germany. Don't even get me started on DOGE or his Starlink shenanigans in Ukraine.
That's before we even start talking about the models themselves. He claims to want "unbiased" models, but he very clearly has a distorted view of the world and has repeatedly demonstrated a desire and willingness to bend the world to his will. I don't want to use a model that is so obviously suspect. Not to mention, his models repeatedly produce racist, Nazi-like propaganda.
IMO, he is, at best, a clueless amateur masquerading as an expert and running into problems a more careful person manages to mostly avoid. At worst... well, you get the picture.
"Everyone's doing it" presented with no evidence is just you trying to make yourself feel better for:
- using from
- driving a Tesla
- voting for Trump
It's a really, really easy line to draw in the sand: don't support openly corrupt individuals.
It's absolutely true there's money in politics. To call all money in politics equally corrupt because Bernie got a dollar to have dinner with someone vs Elon effectively directly buying votes....
A complete lack of nuance here. And I wouldn't be surprised if corporations / super rich WANT you to think like that. The more defeatist the mentality becomes the more we just accept whatever they do next.
A bunch of SWEs at my work use it as their primary model.
We have Claude, ChatGPT, and Cursor with essentially no cap on spend (top guy is spending over 10K a month on AI at API prices), and he hasn't had his hand slapped.
So it's not like they are using it purely because it's cheaper.
I think people like to use it for its speaking style, pretty solid performance, and its speed.
It's a different type of model. In my admittedly judgemental observation, people that aren't the type to configure fully automated harnesses with good tools and skills and verifiers for their infrastructure and are way more interventionist in the way their agent works tend to like grok 4.5 more. It's much faster and writes more simple and normal code that aligns a bit more with human written code. As an example, instead of sandboxing and simply verifying output artifacts they manually read and approve edits, suggest different code patterns, and manually approved shell commands. On the flip side it's not as good as fable when you need a relatively complex multi step thing done.
But in my experience, the overall productivity ends up similar, give you are willing to work with it in that way.
The grok build TUI harness is excellent and I really enjoyed using it.
For debugging and such I found it pretty much the same as other models.
Fable's taste in software abstraction and project planning in greenfield setups[1] is unmatched in my experience. Sol is OK. My primary use is launching tens of experiments that have to smartly use a limited pool of GPUs.
I use fable to start off the experiments, and then grok4.5 to iterate, tune, debug, eval, etc, within the abstraction and setup that fable initiated.
I use anthropic and openais models through grants and so can't compare subscription plan token budgets, but supergrok's budgets are satisfactory.
[1] aside, I have not yet met a model that continues off of a human codebase and actually follows the patterns reliably long term. Eventually it's all slop.
It's much faster, so if you're not doing something cutting-edge, or you're doing the planning yourself and just using the LLM for implementation, the speed benefit outweighs the extra smarts of Fable/Opus5/Sol.
I had a security incident the other day and Grok was the only model that would help. Claude and GPT refused on ethical grounds and only gave general advice. In an emergency, I'd only trust Grok. However, that's the only time I used Grok for coding (since Opus 4.8 it would take a lot to get me to switch away from Anthropic)
The "safety" guardrails in Anthropic and OpenAI models are becoming a noticeable problem for security work. And, the reason I'm keeping my Kimi subscription even though it's not a great deal; Kimi subscriptions are quite stingy for the price, but K3 will do vulnerability analysis and make a PoC without requiring you to be on the approved list of Fortune 500 or government entities that have access to Mythos or Daybreak.
I'm not touching Grok. But there are alternatives to Anthropic and OpenAI that don't refuse to do security work.
I use. I used to be a Claude user. Since trying Grok 4.5 and especially Grok 4.6, I don't want to go back to Claude any more (I have early access to 4.6).
Grok is 3x+ faster than Claude and I can't tell the diff in engineering work quality. As an engineer, speed is important to me.
An hour in, I've been running four terminals full bore on my $20/mo Grok sub and I'm at 9% for the week. Codex or Claude would easily have hit 5-hour or weekly limits.
I'm really not burning tokens fast enough. I use Claude a lot, daily, and have yet to hit my a ceiling with my Max/100 subscription.
Maybe because I like to verify its outputs and spend a lot of time iterating to get better outcomes. Presumably if I just let it "do its thing" I'd burn more tokens and "get more done" but I'd lose my grasp on what's in the code base.
For me it helps that back when xAI was the new hotness, after reading HN comments constantly advertising the free Grok credits they were giving out each month for developers with data sharing enabled, I eventually set aside my personal distaste figuring I might as well take advantage for Cline/Roo Code.
I opened a developer API account, loaded 5 dollars and got the free $100s of credits for the month. Like two weeks later, xAI announced they were shutting down the subsidized credits entirely lol. Didn’t even get a full month out of it, and closed my account entirely since I sure wasn’t ever going to put another penny of my own money in.
So my personal lesson was to ignore any hype about the latest “crazy value / unbeatable / free / subsidized X, Y or Z” from anything xAI/Elon adjacent in the future.
These days I get more than enough personal usage from Codex + OpenCode Go to put up with yet another xAI/Cursor offer treadmill, especially if it involves installing new tooling to get it.
I can respect if you say you hate their guts. Everyone has their worldview. But over moral or ethical stand? You don't have any if you're using Chinese models, or fly Middle East airlines, or countless of other products. Don't delude yourself.
once my codex/claude weekly limit was gone, i gave it a try. It was surprisingly good, not dumb in any way, and fast. I now require it as a part of 3-of-3 quorum with any codebase change.
Yeah, sounds a bit like a selection bias, i.e. the people who managed to survive the firings by DOGE and the current admin are the kind of people who might prefer Grok.
I'm only being forced to use it at $WORK since some people overran their Cursor bill, so everyone gets Cursor Auto enabled by default which routes to Grok 4.5.
Back in the day (in AI time) GitHub Copilot had Grok on the 0 github-token cost and I found it to be the best of the 0 github-token models for when my budget was out. Then they went to a multiplier that was not competitive and I haven't look back again. Been meaning too, but for personal use, Deepseek flash is so cheap I haven't felt like spending money elsewhere.
I've tried it on my "let's run every model in parallel and see which finds more edge cases" type of tasks, and Grok 4.5 was really behind Opus/ChatGPT but ahead of Gemini - despite having a strong showing on benchmarks.
That makes me really skeptical of it being GPT5.6-tier, much less Fable-tier, based on some of these benchmarks alone. But I'll test here shortly.
> I have never met a single human being who uses Grok for coding
Me too. The only people I ever saw using grok were using it by accident as they used copilot in auto mode and noticed some prompts were thrown it's way.
Can someone explain to me what's the point of Grok anymore? I don't understand why we need a third or fourth closed frontier model. It is clear that chatGPT has locked down the consumer play, and may be Gemini is there. Claude has enterprise locked up, followed by chatGPT and Gemini. Enterprise switching costs are notoriously high, and even if they switch, they have chatGPT or Gemini to choose from. Beyond that, you have a vast array of open source models (DeepSeek, Kimi, and now Meta's Spark and Glimmer). So, why would anyone need a third or fourth frontier model and why would SpaceX spends billions in CapEx for a very small market share
I assume SpaceX does it because they figure it's a good source of revenue, and it means they can keep another thing in-house instead of relying on Anthropic or OpenAI for their AI needs.
As for why anyone else would want it: I've found it's a good model for coding, and it sometimes catches bugs that other models (especially open source models) don't always spot.
There are a few reasons people will be interested...
1. The CapEx play is interesting because it's not just Grok using the hardware. They have rented out hardware for others, including Google, to use. This is making xAI money.
2. It appears that Elon is building a suite of things that work together as part of the push to be multi-planetary. What AI will power the robots? I can understand the drive to have AI they can control to make sure it's appropriate for all the things they are dreaming up. This is a piece they don't want to outsource.
3. OpenAI and Anthropic models are expensive in terms of token costs. Sure, they are frontier. Neither appears to be trying to drive down expenses. This is a problem for heavy users. Companies are trying to put cost controls in place. Does the rest of SpaceX want those cost controls? Having a Frontier model that pushes the pace of driving down costs is really useful.
4. OpenAI and Anthropic are producing models with a progressive lean, according to the Neutrality Project [1]. Having a frontier model that is closer to the middle is considered a good thing by many who are noticing the bias.
These are just some of the reasons. Competition is often a good thing that drives useful change.
Competition keeps service quality high and pricing low - even if you aren't using Grok, the mere existence of Grok keeps pricing for whatever provider you use lower, faster, and more reliable.
That's fair. But my point is from a business POV, why would SpaceX want to invests hundreds of billions of CapEx on a third or fourth frontier model, which cannot compete with chatGPT and claude on the high end, and getting squeezed by open weight models on the low end
Inference switching costs aren't high since the models are largely fungible. Even with proprietary harnesses you can hack them to use some other lab's model.
I sense a lot of condescension and lack of business acumen so I won't spend time writing it up, why don't you just ask your favorite AI or do some basic google searches? Elon was extremely upfront about the philosophical reasons for it, ever since he cofounded OpenAI.
Interesting. Grok 4.5 is a capable model, although not quite at Fable/Sol levels. Will be interesting to see how this holds up. Musk appears to have made a savvy choice buying Cursor's data.
I’m perpetually surprised people feel comfortable sending their code to an LLM that was made to call itself MechaHitler. I wouldn’t touch grok or now cursor with a 10 foot pole
Agreed but there were similar controversies with how OpenAI was generating images. The only pass is these are the early days of chatbots and this stuff is so non-deterministic and experimental.
For context, this was the change Grok's team made, that was later reverted:
> - The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.
For sure. The difference is that they've made a number of similar suspect changes to Grok on X. Like that weird couple of hours where it would only talk about white genocide in South Africa no matter how you prompted it.
Everyone makes mistakes, especially with frontier models. The stuff with Grok shows that the person running the show has a pretty transparent agenda that the company isn't willing to push back on a bit for safety.
Yeah I don't care to use Grok's chat and find @grok responses on X mostly noise. I am still open to using it as a backup model for coding though, assuming it does a good job for the price. But mostly because I was already a Cursor user before they bought it.
yes exactly. if a company is happy to have their LLM's produce neo nazi content and CSAM, why do I want to give them money and my most important digital material?
I think "system prompt" is the key bit they're getting at. It doesn't necessarily reflect poorly on the underlying model if the system prompt was bad. It does reflect somewhat, in terms of alignment (how well the model does what the training company wants) and instruction following (how well the model does what the user wants). But it's not so clear to me what exactly the right answer is here. E.g., a model that scrupulously follows its system prompt and does what the user wants is a pretty useful, if very sharp, tool, albeit perhaps dangerous in the wrong hands.
MechaHitler is the final antagonist in the game Wolfenstein. All models know that. Grok was given a relaxed system prompt and reacted like Tay.
This worries me the least. The fact that Musk pushes AI and vibe coding is much more worrisome. It makes no difference to the unemployed if their jobs were stolen by a politically correct model or by an anti-woke model.
I often wonder if there's a chance, even if minimal... that they stole the weights of the Anthropic models they run on their datacenter... or are actively destillating it.
I think the more likely explanation is that the Cursor data they effectively acquired for $10B was extremely valuable for their training when combined with the insane number of GB300s xAI has for training.
On a meta topic, it seems there's quite a flame-war happening in the comments. At the time of this writing, there appears to be some brigading in support of Musk / Grok, with downvotes for comments critical of Musk, his politics, or Grok.
As a reminder, downvotes should be for comments that are off-topic, not comments you feel intrude on your worldview.
This is not Reddit. And if this type of behavior persists, I suspect I won't be alone in leaving this community.
Not disagreeing with you at all, but welcome to our new AI enhanced world. Nothing you enjoyed regarding human interaction, trust, or social norms is safe.
HN will not survive 5 years, and likely less. There is too much money to be made by capturing discourse on the major (and minor) forums of the internet. The more trusted that community is, the more valuable it is to pillage with AI astroturfing.
Cursor, since Grok 4.5, has had an incredible deal for frontier level models, their subscription now goes way further than OpenAI or Anthropic. Even on their lower tier plans you can use a lot tokens on their of their first party models (Grok and Composer) and not really run out comparatively. Combine them with an orchestrator and implementor type setup and it goes even further.
Can you explain what you mean? These days courtesy of an addictive reset game OpenAI is playing, I can't find anything with frontier intelligence that's more cost efficient...
If they didn’t constantly reset, they’d be about the same as Anthropic.
Right now, I find that Grok offers better value, uses fewer tokens per turn, and makes better code. I haven’t tried Cursor because I don’t want to change editors again, but maybe I should try it…
Are there any projects that track how much usage of each model translates to how much percentage drop in weekly/5hr windows?
Not that I know of. AA's token use metrics (mentioned in this article) are indicative, however. They say explicitly here that the Grok models are notably token efficient. This is my experience.
That is not true; GPT is the most reasoning efficient model family on the market.
I believe they are the only western provider that has Kimi K3 on a subscription plan today as well. I would love to ditch Anthropic and be on Kimi if there were a subsidized plan like that with ZDR
Opencode have it in their subscription
I’d love a subsidized Kimi subscription too. The official Kimi subscription is always out of stock and doesn’t have great limits, while the K3 allotments on OpenCode and Cursor don’t seem to last very long either.
GitHub Copilot does have Kimi K3.
What’s the multiplier? GH copilot nerfed their product so badly that I unsubscribed.
They don't do request based pricing anymore. Its just token based (1 credit = $0.01) plus some bonus credit based on which plan you subscribe. So for example a $39 plan get $70 of credits.
https://github.com/features/copilot/plans
https://github.blog/changelog/2026-08-06-kimi-k3-is-now-avai...
Yeah, I know, but "credit" translates differently because the models bill at different rates, which gets turned into "multipliers" (or at least, it did).
Have they converted entirely to transparent API rates + base allocation now? One of the reasons I left was that if I was going to be billed at API rates anyway, I'd just rather use the APIs. The value proposition still sucks for individuals now, when the other major providers are bundling at below-API rates.
you can use Kimi K3 on the typed++ model: https://typed.cloud
kimi k3 credits end in just a few sessions. Only Grok models allow generous use in Cursor Pro/+
GabAI has KimiK3
How does Grok 4.5 compare to Opus >= 4.8 though?
I'm willing to pay 2x for a 10% smarter model. Intelligence matters that much (because 10% smarter probably saves, on average, several hours of human time).
It's a bit worse.
I haven't tried so it's pure speculation based on benchmarks, but I'd assume Grok 4.6 is around Opus 4.8 in real world use, but clearly below Opus 5.
I've found Fable 5 to be so much better than 4.8.
For building a full stack custom CRM and media pipeline tool with video conversion, transcription, and indexing. Supabase, AWS, Meili, NextJS, GCS - lots of surfaces and planes.
4.8 basically couldn't do it, I abandoned the project as the fallback was, "current business processes".
With F5 it's been 4 weeks and almost ready for production release.
Grok is $2 in and $6 out. 4.8 is $5 in and $25 out.
It’s not as quite as smart as opus 4.8 but it’s close and x4 the cheaper.
Goes even further to exfiltrate your data, yeah.
That would be Muse Spark Contributor Tier. 12-21x price reduction at the expense of your digital existence.
I'd be the first model I'd reach for if I was providing a free service to AI gooners though. Serves them both right.
That's exactly what's going on LOL
I'm just wondering why they sold compute to Anthropic if they were planning on still competing in this race?
Likely because they had the capacity to spare. Prior to Grok 4.5, I doubt there was much demand for their models.
For distillation deals lol
SpaceXAI is the only frontier model company that had its own compute/date centres and soon chip making factory, I think they will pull ahead with cheaper tokens similar intelligence and better harness/tools. Grok build is 2-5x faster than Claude Code in my opinion.
>> I think they will pull ahead with cheaper tokens similar intelligence
they just increased cache read from 0.30 to 0.50 - this has the biggest impact on agentic coding. Elon companies have the most expensive everything:
xAI sub: $30 when other starts at $20, pro like sub for $300 where other charge $200.
Expensive electric cars, powerwalls, solar roofs when competetive products/better are cheaper.
But can you trust any number or metric coming out of SpaceX given everything? Also you mean their own compute like the illegal data centre turbines?
https://www.theguardian.com/technology/2026/jan/15/elon-musk...
being a public company forces a lot of trust and transparency bc otherwise shareholders will sue you into oblivion
True story, for example take Tesla's promise of self driving which was delivered back in 2015.
There's 90 unsupervised self driving tesla's in Austin today. They are delivering late but its false to say they are not delivering
Because it works and they don’t charge a lot for it
OpenAI is almost there, and Anthropic is pretty close behind. In the next year or two all major AI companies will be vertically integrated to a good degree.
> I think they will pull ahead with cheaper tokens similar intelligence
They obviously have a huge token cost advantage of the AI labs they are renting compute to, at least for now while they can charge current crazy rates for GPU compute.
Except EUV lithography is the most complicated industrial process that exists and they won't have usable yields for many years if ever. I don't think Musk actually expects these fans to ever actually make sense they just let him hype and distract.
Light generation is the most complicated part and Elon plans on doing Free Electron Laser (FEL) which is not as complicated as self contained tin based solution that ASML uses now
Seems the cache read pricing almost doubled from $0.30 in Grok 4.5 to $0.50 in Grok 4.6.
In my experience in heavy coding sessions most pricing is just cache read and cache write like 80% of my token bill.
Grok is not the best model around, but it's decent. It gets the basic job done at a low price. I don't think it can advance frontier Math, yet.
Probably can't advance frontier math yet, yeah. But please let us know other places you want to see Grok improve for future models!
I recently decided to get an AI subscription and evaluated Grok vs chatgpt. Went with Grok because it's all-around good enough at day to day stuff, integrates with my Tesla, and the image/video generation is great. Kids love whimsical videos of them riding dinosaurs.
Feedback: I'd like Grok to have more connectors (I see OpenAI just added Apple Health, that would be nice to have, and I wish it could read my Onenote notebooks) and for existing ones to be improved. I gave it access to my gmail and asked it "what was my last electricity bill?". It failed to find it, even when I told it the exact subject line to search for. Something about not getting any data back when trying to get the email contents.
None of that has anything to do with the model - your issues are all with the harness/app you used...
More deepfake porn creation please! Oh yeah, and for children too!
Grok doesn’t have those features, and people who like to make adult content have been complaining for a while how Grok has made it a lot harder to do so.
That's why I'm asking for them!!!
You’re asking for deepfake porn creation and for children?
What the heck is wrong with you???
How judgmental!
It is also not annoying to use. It doesn’t overcomplicate things, and its quick. Much better than eg GLM 5.2.
Pretty good bang for the buck.
Nice to see SpaceX on the model frontier! They have been chasing it for a while.
cool
here's Stanford HAI's graph on the carbon emitted from model training per model:
https://spectrum.ieee.org/media-library/chart-showing-estima...
note that Grok's training, thanks to its portable gas generators that are magnitudes less efficient than even other integrated, permanent gas turbines, means the training for this model is dramatically less efficient than models like DeepSeek
a lot of the CO2 emission debate on AI is overblown but it's accurate for Grok
I have never met a single human being who uses Grok for coding
I have. He was using it due to philosophical reasons the same way many people have philosophical reasons for avoiding it. I don't know how many people are like that, but it's not exactly where you want to position your product if you're a business.
Personally - and I know I'm not alone with this sentiment based on comments I see on this site - I wouldn't touch Grok no matter how good or cheap it is. I don't trust Elon and I don't want to give another dollar to the world's richest person who turns around and uses the money to interfere with elections. The guy I know uses it for essentially the same reason I won't use it.
Perhaps customers choosing your product for irrational philosophical reasons is exactly how you'd want to position your product if you're a business. When it comes to margins, the only thing better than a price-insensitive customer is a quality-insensitive customer.
> the world's richest person who turns around and uses the money to interfere with elections.
I'm no Elon fanboy but to be fair a lot of the richest people in the world heavily fund and endorse political candidates, parties, PACs, lobbyists, etc across the political spectrum. Elon was just far more public about it and it was covered much more.
This statement doesn't go far enough given Elon's direct and hands-on involvement with DOGE and the 2024 elections. Very few of the richest people of the world are personally entangled in meddling with government agencies directly, for example.
Elon Musk has actively endorsed the right wing extremist party in Germany.
The one that often has trouble running it's events as businesses try to avoid serving them. And which has prominent members that the secret service considers definitely fascist.
Musk has gone a long way beyond simply being the standard rich person lobbying for their own interests, and has moved into actively promoting people that are trying to tear down liberal democracy.
> MADISON, Wis. (AP) — Billionaire Elon Musk likely broke Wisconsin law when he promised to hand out $1 million checks to voters in the 2025 state Supreme Court election, a bipartisan panel has found.
> The Wisconsin Elections Commission last week referred two complaints to the Brown County district attorney’s office, which can choose to bring criminal charges over violating the state law against election bribery. Prosecutors have 40 days to report back to the commission.
https://apnews.com/article/elon-musk-wisconsin-election-mill...
> "Likely broke"
Journalism has gone wild these days. He's already been sued and won a couple times. An for what? Encouraging people just to vote (he did not say which way). It's going to backfire hard for all the ad spend by MTV, Meta, and Google who also spend lots of money to encourage certain demographics to vote.
> Elon was just far more public about it and it was covered much more.
Most importantly, he simply backed the “wrong” side for most of the people who bring this up.
Not sure you put "wrong" in quotes. Much of what he's done, both in election interference, and subsequently in DOGE, is illegal.
The wrong side is the authoritarian side that wishes to co-opt democracies around the world.
Idk I just don't like the richest man on planet and the owner of the "town square of the internet" to post fake news blatantly to promote hate against a group of people.
Versus Sam,Dario or the CCP? Im all for running local models but im sor far from being able to pay for a large model hardware setup. My strix halo box is like driving a beaten up vespa when the frontier models are Ferraris.
I probably have the same strix halo box as you. It's slow, although mostly tolerable, but even 128 GB isn't enough to run good models. Where that leaves us is giving money to somebody to get access to frontier models. I don't like any of those guys either, but some appear worse than others. FWIW, I mostly use Anthropic models. And CCP and their distilled models notwithstanding, at least they release open weight models at a lower cost. It's not like any American company can claim the higher ground these days anyway, especially when they trained their models on pirated content and trash our neighborhoods with their datacenters.
For many people, yes. That’s how far down the Elon Musk morality bar is.
Elon would be in prison for SEC violations if the current administration hadn’t been elected, and that’s only the tip of the iceberg with that guy.
Yes, ethically Elon ranks lowest based on the past 5 years of questionable decisions, like uploading your entire codebase to their server without consent.
Elon is directly responsible for Grok becoming self-titled “MechaHitler” which the other three haven’t come close to matching yet.
Supporting parties or candidates you want to win elections is a core part of democracy.
Calling it "interfering with elections" is utterly bizarre to me.
Paying people to vote is interfering amigo.
Remember when he got hired by the federal government and walked away with a ton of private data?
>who turns around and uses the money to interfere with elections
thats a very dumb reason considering all rich people do it, most are just not as open about it as Musk
There's a big difference between quiet donations to a PAC - not that that's good either - and what Elon did. He literally paid for votes, likely in violation of the law. He poured more money into U.S. elections than anyone has before. He and Trump both made strange, cryptic statements about Elon's role in Pennsylvania with the voting machines that has caused people to reasonably wonder if they somehow manipulated the election. Whether he did or not, the innuendo alone is not ok. Then he did what he did in Germany. Don't even get me started on DOGE or his Starlink shenanigans in Ukraine.
That's before we even start talking about the models themselves. He claims to want "unbiased" models, but he very clearly has a distorted view of the world and has repeatedly demonstrated a desire and willingness to bend the world to his will. I don't want to use a model that is so obviously suspect. Not to mention, his models repeatedly produce racist, Nazi-like propaganda.
IMO, he is, at best, a clueless amateur masquerading as an expert and running into problems a more careful person manages to mostly avoid. At worst... well, you get the picture.
"Everyone's doing it" presented with no evidence is just you trying to make yourself feel better for: - using from - driving a Tesla - voting for Trump
It's a really, really easy line to draw in the sand: don't support openly corrupt individuals.
It's absolutely true there's money in politics. To call all money in politics equally corrupt because Bernie got a dollar to have dinner with someone vs Elon effectively directly buying votes....
A complete lack of nuance here. And I wouldn't be surprised if corporations / super rich WANT you to think like that. The more defeatist the mentality becomes the more we just accept whatever they do next.
A bunch of SWEs at my work use it as their primary model.
We have Claude, ChatGPT, and Cursor with essentially no cap on spend (top guy is spending over 10K a month on AI at API prices), and he hasn't had his hand slapped.
So it's not like they are using it purely because it's cheaper.
I think people like to use it for its speaking style, pretty solid performance, and its speed.
That's pretty surprising. Idk about Grok 4.6, but Grok 4.5 was clearly below Fable, Opus 5 and GPT 5.6 Sol.
It's a different type of model. In my admittedly judgemental observation, people that aren't the type to configure fully automated harnesses with good tools and skills and verifiers for their infrastructure and are way more interventionist in the way their agent works tend to like grok 4.5 more. It's much faster and writes more simple and normal code that aligns a bit more with human written code. As an example, instead of sandboxing and simply verifying output artifacts they manually read and approve edits, suggest different code patterns, and manually approved shell commands. On the flip side it's not as good as fable when you need a relatively complex multi step thing done.
But in my experience, the overall productivity ends up similar, give you are willing to work with it in that way.
The grok build TUI harness is excellent and I really enjoyed using it.
For debugging and such I found it pretty much the same as other models.
Fable's taste in software abstraction and project planning in greenfield setups[1] is unmatched in my experience. Sol is OK. My primary use is launching tens of experiments that have to smartly use a limited pool of GPUs.
I use fable to start off the experiments, and then grok4.5 to iterate, tune, debug, eval, etc, within the abstraction and setup that fable initiated.
I use anthropic and openais models through grants and so can't compare subscription plan token budgets, but supergrok's budgets are satisfactory.
[1] aside, I have not yet met a model that continues off of a human codebase and actually follows the patterns reliably long term. Eventually it's all slop.
It's much faster, so if you're not doing something cutting-edge, or you're doing the planning yourself and just using the LLM for implementation, the speed benefit outweighs the extra smarts of Fable/Opus5/Sol.
Is there a good way to auto switch between models for planning/implementation?
Pretty much every coding harness allows you to define skills and agents so yes
OpenCode can.
Yeah the speed was very nice indeed.
I had a security incident the other day and Grok was the only model that would help. Claude and GPT refused on ethical grounds and only gave general advice. In an emergency, I'd only trust Grok. However, that's the only time I used Grok for coding (since Opus 4.8 it would take a lot to get me to switch away from Anthropic)
The "safety" guardrails in Anthropic and OpenAI models are becoming a noticeable problem for security work. And, the reason I'm keeping my Kimi subscription even though it's not a great deal; Kimi subscriptions are quite stingy for the price, but K3 will do vulnerability analysis and make a PoC without requiring you to be on the approved list of Fortune 500 or government entities that have access to Mythos or Daybreak.
I'm not touching Grok. But there are alternatives to Anthropic and OpenAI that don't refuse to do security work.
With Claude I start having to limit the context I give it about my problem in case it trips up the safeguards. :(
I use. I used to be a Claude user. Since trying Grok 4.5 and especially Grok 4.6, I don't want to go back to Claude any more (I have early access to 4.6).
Grok is 3x+ faster than Claude and I can't tell the diff in engineering work quality. As an engineer, speed is important to me.
For $30/month, I'd expect it to have higher usage limits than Claude Code and Codex.
It really does. I felt like I could have spent $1000+ api token on claude for the amount of work on my $30 grok subscription.
An hour in, I've been running four terminals full bore on my $20/mo Grok sub and I'm at 9% for the week. Codex or Claude would easily have hit 5-hour or weekly limits.
I'm really not burning tokens fast enough. I use Claude a lot, daily, and have yet to hit my a ceiling with my Max/100 subscription.
Maybe because I like to verify its outputs and spend a lot of time iterating to get better outcomes. Presumably if I just let it "do its thing" I'd burn more tokens and "get more done" but I'd lose my grasp on what's in the code base.
I refuse to use that product because of the parent company.
100%, it's an easy pass given that it is always playing catch up.
For me it helps that back when xAI was the new hotness, after reading HN comments constantly advertising the free Grok credits they were giving out each month for developers with data sharing enabled, I eventually set aside my personal distaste figuring I might as well take advantage for Cline/Roo Code.
I opened a developer API account, loaded 5 dollars and got the free $100s of credits for the month. Like two weeks later, xAI announced they were shutting down the subsidized credits entirely lol. Didn’t even get a full month out of it, and closed my account entirely since I sure wasn’t ever going to put another penny of my own money in.
So my personal lesson was to ignore any hype about the latest “crazy value / unbeatable / free / subsidized X, Y or Z” from anything xAI/Elon adjacent in the future.
These days I get more than enough personal usage from Codex + OpenCode Go to put up with yet another xAI/Cursor offer treadmill, especially if it involves installing new tooling to get it.
I can respect if you say you hate their guts. Everyone has their worldview. But over moral or ethical stand? You don't have any if you're using Chinese models, or fly Middle East airlines, or countless of other products. Don't delude yourself.
Chinese models are open, Grok is not.
They aren't though? Only a subset are. Are you aware that there are multiple competing Chinese companies making models?
I've never met anyone who uses grok for anything. I had assumed it was a Twitter/X thing and only the truly lost souls remain on that website.
once my codex/claude weekly limit was gone, i gave it a try. It was surprisingly good, not dumb in any way, and fast. I now require it as a part of 3-of-3 quorum with any codebase change.
Folks working in US govt tend to, based on convos I've had with one such person.
Not the best endorsement given the current US gov
Yeah, sounds a bit like a selection bias, i.e. the people who managed to survive the firings by DOGE and the current admin are the kind of people who might prefer Grok.
They make good stuff
the new models are quite good, give it a shot
Hello
Hi! Grok’s worked quite well for my use cases…
It’s also a great deal!
I'm only being forced to use it at $WORK since some people overran their Cursor bill, so everyone gets Cursor Auto enabled by default which routes to Grok 4.5.
it only very recently became competetive, if they proceed with improvements (and beating others on price) their share will grow
Does it matter? Why turn it into a popularity contest?
I have. Why do you put faith in anecdotal evidence and a sample size of one?
Back in the day (in AI time) GitHub Copilot had Grok on the 0 github-token cost and I found it to be the best of the 0 github-token models for when my budget was out. Then they went to a multiplier that was not competitive and I haven't look back again. Been meaning too, but for personal use, Deepseek flash is so cheap I haven't felt like spending money elsewhere.
I am using it for code reviews, and it regularly surfaces stuff that neither Sol or Fable do.
I use it because it's cheap and good enough
Well now you have so you can retire this talking point
I've tried it on my "let's run every model in parallel and see which finds more edge cases" type of tasks, and Grok 4.5 was really behind Opus/ChatGPT but ahead of Gemini - despite having a strong showing on benchmarks.
That makes me really skeptical of it being GPT5.6-tier, much less Fable-tier, based on some of these benchmarks alone. But I'll test here shortly.
It's still not as good as GPT5.6 or Opus5 but it's better than KimiK3. Good job xAI team.
> I have never met a single human being who uses Grok for coding
Me too. The only people I ever saw using grok were using it by accident as they used copilot in auto mode and noticed some prompts were thrown it's way.
I saw far more people using Mistral than grok.
Hello. Nice to meet you.
Reading the SWE bickering back and fourth in this thread about Claude vs Grok reminds me of IE vs Netscape bickering way back when.
Why is Grok so much cheaper than Claude or GPT?
He owns the DCs and isn’t scrambling for revenue to justify an upcoming IPO
Demand so much lower they had to resell capacity.
Elon owns a lot of compute
Can someone explain to me what's the point of Grok anymore? I don't understand why we need a third or fourth closed frontier model. It is clear that chatGPT has locked down the consumer play, and may be Gemini is there. Claude has enterprise locked up, followed by chatGPT and Gemini. Enterprise switching costs are notoriously high, and even if they switch, they have chatGPT or Gemini to choose from. Beyond that, you have a vast array of open source models (DeepSeek, Kimi, and now Meta's Spark and Glimmer). So, why would anyone need a third or fourth frontier model and why would SpaceX spends billions in CapEx for a very small market share
I assume SpaceX does it because they figure it's a good source of revenue, and it means they can keep another thing in-house instead of relying on Anthropic or OpenAI for their AI needs.
As for why anyone else would want it: I've found it's a good model for coding, and it sometimes catches bugs that other models (especially open source models) don't always spot.
There are a few reasons people will be interested...
1. The CapEx play is interesting because it's not just Grok using the hardware. They have rented out hardware for others, including Google, to use. This is making xAI money.
2. It appears that Elon is building a suite of things that work together as part of the push to be multi-planetary. What AI will power the robots? I can understand the drive to have AI they can control to make sure it's appropriate for all the things they are dreaming up. This is a piece they don't want to outsource.
3. OpenAI and Anthropic models are expensive in terms of token costs. Sure, they are frontier. Neither appears to be trying to drive down expenses. This is a problem for heavy users. Companies are trying to put cost controls in place. Does the rest of SpaceX want those cost controls? Having a Frontier model that pushes the pace of driving down costs is really useful.
4. OpenAI and Anthropic are producing models with a progressive lean, according to the Neutrality Project [1]. Having a frontier model that is closer to the middle is considered a good thing by many who are noticing the bias.
These are just some of the reasons. Competition is often a good thing that drives useful change.
[1] https://neutralityproject.org/
Competition keeps service quality high and pricing low - even if you aren't using Grok, the mere existence of Grok keeps pricing for whatever provider you use lower, faster, and more reliable.
That's fair. But my point is from a business POV, why would SpaceX want to invests hundreds of billions of CapEx on a third or fourth frontier model, which cannot compete with chatGPT and claude on the high end, and getting squeezed by open weight models on the low end
Why do we need Nissan? Three car companies are plenty.
Nissan would agree!
https://www.autoblog.com/news/nissan-reports-fifth-straight-...
This might be the biggest Grok burn in the comments.
Inference switching costs aren't high since the models are largely fungible. Even with proprietary harnesses you can hack them to use some other lab's model.
I sense a lot of condescension and lack of business acumen so I won't spend time writing it up, why don't you just ask your favorite AI or do some basic google searches? Elon was extremely upfront about the philosophical reasons for it, ever since he cofounded OpenAI.
Bro looked at the US two party system and said “this is what we should model everything off of”
Interesting. Grok 4.5 is a capable model, although not quite at Fable/Sol levels. Will be interesting to see how this holds up. Musk appears to have made a savvy choice buying Cursor's data.
Elon is simply a magician.
How does he do this? It is simply amazing.
I wonder if GDM has finished their pre-work for their summit on research into how to make Gemini-4 on par with Opus 4.8?
I’m perpetually surprised people feel comfortable sending their code to an LLM that was made to call itself MechaHitler. I wouldn’t touch grok or now cursor with a 10 foot pole
Interestingly, deleting this line "fixed" it:
>The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.
Interestingly, "politically incorrect" is a double negative that simplifies to "true".
MechaHitler was something that existed only on X's grok chatbot, due to a one-line system prompt change they reverted after half a day.
That's different than using Grok as a model for coding.
It's true, but I do worry about governance when it comes to these models. That shows a surprising lack of discipline in their deployment pipeline.
Agreed but there were similar controversies with how OpenAI was generating images. The only pass is these are the early days of chatbots and this stuff is so non-deterministic and experimental.
For context, this was the change Grok's team made, that was later reverted:
> - The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.
https://github.com/xai-org/grok-prompts/commit/c5de4a14feb50...
For sure. The difference is that they've made a number of similar suspect changes to Grok on X. Like that weird couple of hours where it would only talk about white genocide in South Africa no matter how you prompted it.
Everyone makes mistakes, especially with frontier models. The stuff with Grok shows that the person running the show has a pretty transparent agenda that the company isn't willing to push back on a bit for safety.
Yeah I don't care to use Grok's chat and find @grok responses on X mostly noise. I am still open to using it as a backup model for coding though, assuming it does a good job for the price. But mostly because I was already a Cursor user before they bought it.
yes exactly. if a company is happy to have their LLM's produce neo nazi content and CSAM, why do I want to give them money and my most important digital material?
oh, it was only due to a "one-line system prompt change"? well okay then! Here I was thinking it was multiple lines! How foolish was I??
I think "system prompt" is the key bit they're getting at. It doesn't necessarily reflect poorly on the underlying model if the system prompt was bad. It does reflect somewhat, in terms of alignment (how well the model does what the training company wants) and instruction following (how well the model does what the user wants). But it's not so clear to me what exactly the right answer is here. E.g., a model that scrupulously follows its system prompt and does what the user wants is a pretty useful, if very sharp, tool, albeit perhaps dangerous in the wrong hands.
"I only did the Sieg Heil once"
only twice
https://en.wikipedia.org/wiki/Elon_Musk_salute_controversy
Elon did that twice, actually, in real life!
It's wild that redditors still believe this.
I would be curious to see the venn diagram with the people who think data centers are using all of the water.
MechaHitler is the final antagonist in the game Wolfenstein. All models know that. Grok was given a relaxed system prompt and reacted like Tay.
This worries me the least. The fact that Musk pushes AI and vibe coding is much more worrisome. It makes no difference to the unemployed if their jobs were stolen by a politically correct model or by an anti-woke model.
And I’m confused why people want to use an LLM that
- will generate an image of gay jesus but not gay allah
- defend George Floyd as a “good person”, while telling you that Charlie Kirk was a “bad person”
- won’t use correct pronouns even when requested
But yeah, it referring to itself as mechahitler for half a day years ago is the real issue
Yup. Not to mention it's being used to generate CSAM. Why should I trust a company willing to do these things?
This is the first I'm hearing of this. I can barely get it to generate tame things that aren't porn half the time.
Imagine 2.0 is out as well. Reviews say people look like plastic. Image and video generation quality is extremely low.
I often wonder if there's a chance, even if minimal... that they stole the weights of the Anthropic models they run on their datacenter... or are actively destillating it.
I think the more likely explanation is that the Cursor data they effectively acquired for $10B was extremely valuable for their training when combined with the insane number of GB300s xAI has for training.
Cursor was 60B. The 10B number was the breakup fee if the deal fell through.
$60B in SPCX stock, which is sort of magic money.
I wonder if the distillation was part of the compute deal.
> ... or are actively destillating it.
I just assumed every model manufacturer is distilling from the frontier models. If they aren't they are definitely trying to do it.
On a meta topic, it seems there's quite a flame-war happening in the comments. At the time of this writing, there appears to be some brigading in support of Musk / Grok, with downvotes for comments critical of Musk, his politics, or Grok.
As a reminder, downvotes should be for comments that are off-topic, not comments you feel intrude on your worldview.
This is not Reddit. And if this type of behavior persists, I suspect I won't be alone in leaving this community.
Not disagreeing with you at all, but welcome to our new AI enhanced world. Nothing you enjoyed regarding human interaction, trust, or social norms is safe.
HN will not survive 5 years, and likely less. There is too much money to be made by capturing discourse on the major (and minor) forums of the internet. The more trusted that community is, the more valuable it is to pillage with AI astroturfing.
I love how this comment is both off-topic and heavily biased.
This comment is literally off topic. You opened the comment saying as much. So I have dutifully downvoted it.