A follow-up rant arguing that Gemini 4 Carbon is not just slow but unusable for agentic coding, with the latency data to back it up.
Posted
2 days ago
Duration
Format
Talking Head
comedic-rant
Views
146.6K
2.8K likes
57 · 43
Big Idea
The argument in one line.
Gemini 4 Carbon's real-world agentic speed is about five times worse than benchmarks suggest, because an 11-second wait before each response compounds across every step of a multi-turn task.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You're a developer picking a default coding model and want real agentic-work latency numbers instead of single-request benchmark charts.
You use Claude Code, Codex, or similar agent tools and want to understand why time to first token matters more than tokens per second.
You follow the AI lab race and want an inside-the-ring view of how a major lab responds to public criticism.
SKIP IF…
You want a neutral, measured model review. This is an openly angry, opinion-heavy rant with profanity.
You only use models for single-shot chat questions, where per-step latency compounding barely matters.
TL;DR
The full version, fast.
The creator argues his earlier Gemini 4 Carbon review was too soft. Auditing real work, he found it consistently about five times slower than other models, and the cause is not token speed but an 11-second delay between a request hitting Google's servers and the model starting to act. That delay compounds across every step of an agentic turn, and the model also takes more steps and re-reads files 8 to 16 times. Caching is unreliable, and the Vertex hosting and model naming cause extra friction. He concludes the model is not worth using at its price, and urges Google to fix its infrastructure.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Benchmarks that time a single request miss the real cost of agentic coding, where every tool call is another round trip and the delay compounds.
Slow tokens per second, weak parallel tool calls, and thin compute might each cost you 30 to 50 percent. A 5x gap needs a problem that hits before generation even starts.
An 11-second wait before the first action, repeated at every step of a turn, can make a model take longer to start responding than rivals take to finish.
When a tool is slow enough that you tell your agents to give up waiting on it, your own measurements are already flattering it.
A model that reads the same file 8 to 16 times burns both time and tokens, so step count matters as much as per-step speed.
Prompt caching that only hits about 70 percent of the time after you just wrote to it makes cost and latency unpredictable.
Putting reasoning effort in the model name instead of a separate field breaks analytics and tooling that expect one model per slug.
Moving inference to a persistent WebSocket avoids re-sending the whole history each step, which is why Codex once felt faster than Claude.
A model that merely fits cleanly into your existing tools saves more daily friction than one that scores slightly higher on a leaderboard.
Public criticism works best when it is specific and reproducible, and Azure's slow first-token issue was fixed only after sustained public pressure.
Takeaway
Why per-step latency decides agent speed
WHAT TO LEARN
In agentic work, a delay before the first token is paid on every step, so a modest-looking wait can make a model several times slower in practice.
Single-request speed charts mislead for agent work, because every tool call is another round trip and the waits stack across the whole turn.
Audit your own real tasks rather than trusting published numbers, and watch for ways your setup quietly flatters a slow tool, like telling agents to abandon it.
A 5x slowdown is rarely explained by token speed alone. Look at the wait before generation starts, which here was over 11 seconds per step.
Step count matters: a model that re-reads files 8 to 16 times adds time and cost on top of a slow start.
Unreliable caching and non-standard model naming add hidden friction, so judge a model by how cleanly it fits your tools, not only by its raw intelligence.
Specific, reproducible public feedback tends to get infrastructure problems fixed, so share concrete data when you hit a wall with a provider.
Glossary
Terms worth knowing.
Time to first token
How long you wait between sending a request and seeing the model start to respond. It matters most in multi-step agent work, where it is paid on every step.
Pre-fill
The stage where a model reads and processes the whole prompt and history before it can generate its first output token. Longer histories make it slower.
Prompt caching
Reusing already-processed prompt content from earlier requests so the model skips reprocessing it, cutting both latency and cost when it hits.
Agentic turn
A full unit of agent work made of many model calls and tool calls in sequence, as opposed to one question and one answer.
Vertex
Google Cloud's hosted platform for serving its models to developers through an API.
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
As many of you may have noticed, I just published a video about Gemini 4 Carbon, and despite my plans to be really negative towards the model, I found myself strangely okay with it. Not because it writes great code, or that it's good at 3D, or anything other than the fact that it didn't seem like it was going to restrict me too much.
And that was a real, valuable aspect of it, and I really just didn't find it worth using otherwise. I did my best to cover the facts of the model, how much it costs, how fast it really is, what it looks like to use in real world projects. And I was not harsh enough.
This hurts me a bit because I, as I've mentioned, like working with the DeepMind guys. For the most part, they're really upfront, they're really awesome, and they try hard to fix problems. And that's why we're making this video.
Because I was wrong about the speed gap. I had mentioned the fact that 4 Carbon Low was slower than pretty much every other model in history. I was not aware of the fact that it was five times slower.
As many of you guys know, I don't watch my threads when they run very often. I fire them off and then I look when they're done. So I took the time to audit my real world usage of all of these models.
And in like for like work, Gemini was consistently five times slower than anything fucking else. However bad you think this is, it's worse. And I am going to crash out so much about this because it shows on a fundamental level.
that Google does not give a single flying fuck about anyone using their inference, and until these things are fixed, you should not take them seriously as a company. And I hate to be this strict about it, but it is my responsibility to, because I feel like I screwed up here. And I also, to be fair, feel like everybody else is not being real about this either.
People sincerely think the Gemini models are a reasonable alternative to other things. I can't respect you if you pick Gemini right now, because it is that bad. and i am sorry to my friends there i hope you use this video as leverage both to get these things fixed and also in interviews for jobs at companies that actually give a because i am so goddamn done and i'm going to explain in detail the levels of incompetence specifically on the hosting side for gemini After a quick word for today's sponsor.
Back in my day, building a mobile app required a whole dedicated team or two even, one for iOS and one for Android. Getting a whole of this right was really hard. And even with awesome tools like React Native, there were still so many holes you could fall into that made it nearly impossible to get right.
Believe me, I have experience here. I've built a lot of apps with and without React Native. The number of pitfalls you can run into even as an expert is obnoxious.
And your agents might be great, but they're going to hit these even faster and cause even more. problems i've been there it was rough and i'm very thankful that today's sponsor is here to help out infinite red are some of the industry leaders in react native expertise but they go further than that they're more than happy to dive into swift kotlin objective c or even java if necessary if that's what gets the best performance for the thing they're working on they've been helping companies set up their code bases for success in react native forever and it turns out those same things are really good for agents too if you make your code base stupid proof so that engineers who are less experienced with native can make good code Turns out the agents can handle it even better.
And that's what they're here to do, to set you up for success. You go hire a bunch of expensive, dedicated mobile teams, like I'm sure lots of companies have, or you can do this thing that these small businesses like Microsoft or Domino's or Realtor .com has done. Wait, those are big companies.
Yeah. Turns out making good mobile apps is hard enough that even big businesses are relying on Infinite Red to get things right. Whether you're trying to outsource the development of your mobile app, or you're trying to level up your web team to make a better experience on mobile, or you just want to make sure your agents are writing good code, make sure your apps are built right at soydiv .link slash infinite.
I will apologize in advance for how blunt I am being and also for how blunt I have been. I cannot respect the engineering opinions of anyone who thinks this model is usable. for real world engineering work outside of the fact that when you do cyber security stuff with it it's slightly less likely to flag you there is literally no reason otherwise to use this model right now and i have lost respect for a handful of engineers that i thought knew better that are publicly claiming it's good for day -to -day work i am partially guilty here myself i did not think the gap was this big i did not think using it was this miserable and of course the subscription subsidy side it is so disgustingly expensive too especially starting december 1st when they double the price yeah this model is going to double in price in three weeks what am i crashing out about right now let's go through this as i mentioned before a lot of friends at google that have been doing a very good job handling my crash outs getting things fixed where they can and really making things better i know for a fact that they are currently all hands on deck for the speed stuff that i'm complaining about but i need you guys to understand how bad it is in my previous video i showed that when you don't use max effort for anything because you shouldn't max effort is bad
Gemini 4 Carbon is the slowest modern model by a lot. There is no way you can use Opus where it takes longer than Gemini 4 Carbon. And Fable 5 on X High is slightly slower than Gemini 4 Carbon High, but on High it is faster than every single thing Gemini 4 Carbon has completed.
This is wrong and misleading, because this measures from the time the request officially is hit and accepted by the server on Google's side. to when the response comes in, and a lot of the tasks for artificial analysis are not agentic, so they're not multi -turn. Once you get into multi -step turns, which is, guess what, how every single agentic coding tool works, this fucking crumbles so fast.
As I mentioned before, once I started analyzing my data, Gemini 4 Carbon was consistently five times slower in day -to -day work, but this number is actually worse than it seems. I had nine separate times that Carbon was brought in as a judge for work, And you'll notice after the first three, something changed.
It started to say, stopped by the parent a whole bunch. Do you know why that happened? It was so slow that I went and updated my skill to tell my agents, by the way, it looks like carbon's pretty slow.
If you're stuck waiting for it, but you have work to fix, don't bother, just stop it and get back to work. If I hadn't done that, the numbers would have been worse. I accidentally gave them an advantage here.
So not only is it way worse than I thought it was in my first video, it's way worse than I'm showing you now. I don't know how much worse it is because I'm not willing to slow down my day -to -day work enough to properly measure it. I will never learn how much time it takes to hand feed a turtle a full head of lettuce because I'm a busy human being that's not going to sit there and fucking do it.
It's insane. It's so bad. It's so fucking bad.
And it gets so much worse when you see the reasons why. you might think oh it's the tps that's slow oh they're not giving it enough compute oh maybe it's not calling tool calls in parallel properly all those things are true and all of those would make it like 30 to 50 slower you don't get 5x slower by doing tool calls okay you can get 5x lower by doing tool calls wrong they're not doing them that wrong they're close the problem comes before the request even lands this model consistently takes over 11 seconds for when the request hits their server to when it starts doing something.
11 seconds. Let's go through this together in real time. I'm just going to sit here, and we can experience the zen of Gemini 4 Carbon together.
Still going.
Okay, cool. Two of three are gone.
Fable ran a slow tool call in this demo, but it's still finished. And now Gemini's done as well. This was for one step, by the way.
This was not for a proper full turn. This is one step. When you combine this with the fact that Gemini takes more steps to complete tasks, and you have to compound this additional 11 seconds per 20 seconds on every single request, the result isn't just embarrassing.
It's actually unusable. And as I tried to use Gemini more for real work, when I realized I had undercounted its intelligence, I realized I had over accounted my own because it made me feel so fucking stupid to ever pick a model that is this miserable to use. So you might be thinking, as my chat already is, did nobody notice this?
How did this happen? I have seen this happen once before. Is it the small startup that does a little bit of AI stuff?
You might've heard of them. They're named Microsoft. Azure had a similar problem with OpenAI models, where most of the time they would get you the first token in 10 to 15 seconds, but occasionally it would take 100 to 200 seconds, which meant it was the most unreliable, unusable garbage in the fucking world.
I crashed out at them so hard that I had execs hitting me up and they ended up fixing it. Yeah, the million dollar incident. I have a whole video about that one.
I was proud as fuck of that one. and that's why i'm here now the difference is i have nothing to lose because nobody likes gemini that time i had a million dollars on the line this time i'm doing this for the love of the game so why am i talking about this now why am i bothering with another video about gemini as i mentioned before there's a lot of people at deep mind that i actually preferred almost every other lab like i would say the lab i feel like listens to me the best and is the most likely to actually care and try to fix things is deep mind they have been pretty good overall at responding to my crash outs as such i am sorry for this video to my friends there this hopefully will be a useful tool either as leverage you can use in the internal meetings to get these things fixed because they're embarrassing or if that doesn't work which i'll be real it very well might not you can use this video as the greatest letter of recommendation possible when you go to get a job at a real lab because at this point i will not refer to deep mind as a real lab because they are not making technology that you can really use
And as much as I respect most of the people I interact with and talk to at DeepMind, I'm going to get personal a bit because I do not respect Varun. This is the guy that stole a billion dollars from his own team in order to quit the company he started and go to DeepMind with all of the tech. I've never seen...
And it's somebody who cares a lot about their team who has... lost a lot of money in order to do the right thing for their team who could have been bought out personally for a large amount of money but chose not to because i want my team to win as well mark and i are 50 50 split on equity for t3 tools and code and all of that and julius is not far behind us the only people who made any money from windsurf are him and like two of his friends they left 130 person company they started the three of them left it just out to die and all three of them took their money i'm going to be so real with y 'all I'm choosing to get more personal here because I want to see if Rune can handle pressure.
I want him to know that I think he's a spineless prick. I want him to know that I think he's a bad leader. I want him to know that the way he responded to me in this scenario is unacceptable.
First off, like saying this is high praise instead of acknowledging the issues that I described, which mind you, like every other person I talked to at DeepMind has been awesome about this. They've been DMing me all day for the last two days, begging me for feedback. trying their fucking hardest to make the models better and the tools more usable.
He drops a laugh emoji and some annoying but useful advice because I crashed out about my email always being visible in their CLI. Cool. Awesome.
Fine. I asked them to make it a config. Said awesome.
Around this point, I had started digging deeper into the speed issues because I was using it more and it was annoying me. I had not dug deep enough to know exactly what was causing the problems, but I assumed because I had just done an analysis of how tool calls were working with Gemini. i learned it was doing things in parallel i didn't think it was and it was so i was confused how it was still so slow so i recommended a video i had done in the past which is about open ai moving to web sockets for inference in codecs the reason they did that and i've talked about this before watch that video if you're curious the tldr is that the model needs your whole history so when you make a request for the next turn or the next step in a turn it would send the whole history up which is just so much latency sending the whole thing up waiting for it to be parsed matched to a catch pre -filled and then loaded in so it could start responding every time a tool call happened it was just a ton of overhead and one of the reasons codex used to feel so much faster than claude was because they put this layer in front of the inference to connect over a web socket and send less data back and forth and keep it warm it was awesome it was so much faster
I don't know how Anthropic caught up without doing the WebSocket thing, but they did, and it's great. I think a lot of it is cash changes. Regardless, ready for the response?
Time to first token is also just high because pre -fill is not cheap. Frowny face. It's not just network transport.
He's right here. That's correct. Let me quantify, but I'm not sure that would immediately solve this.
And to his credit, he's right. This probably wouldn't immediately solve it. But I was annoyed enough about this that I chose to dig in more.
And that's when I found these much, much crazier numbers. Gemini 4 Carbon takes more time to start responding than 6 -1 Sol takes to finish responding in most scenarios. Insanity.
And then there's the funny Gemini -isms here, like the fact that it on average reads a file 8 -16 times. Hilarious. I don't think you guys properly understand how far behind they are right now.
It's not like 1st and 2nd and 3rd. It's first and second, then like out in like a different school somewhere. It's like the two best students in college and then somebody who has held back four times in kindergarten and dropped out.
It's so bad. And it's already fucking cached. I've been saying this for a bit and I did bring this up in my previous Gemini video.
They are so goddamn bad at caching. It's insane. It's insane.
They're the only lab where you write to cache and then read from cache and it's a 30 % chance it doesn't read from the cache you just wrote. it'll read from something it might be a cache from 30 minutes ago it might just fail to read the cache entirely it's so bad it's so bad this is what i get for actually trying to use the model because i found something it could do which gave me a reason for the first time in a year and a half plus to actually try a gemini model for day -to -day work and i feel like i've been going insane for the past 24 hours as i've tried my hardest to use this model The problem with Google isn't just that Gemini models are bad to work with.
It's that everything around them is crumbling under the weight of this company that doesn't know what the fuck they are doing. It's insane. And you'll never guess who replied to this tweet.
Give you a hint. I don't like him. Not what we're seeing on our workloads.
I got some advice for you, Varun. DM me in the future. I do not like how you engage with me publicly and it makes both of us look bad.
The best thing for both of us here. Is for you to accept that I know what the fuck I'm doing. Because I know you do.
You are clearly very smart. You know your shit. You're just a bad leader.
Just DM me in the future please. Ideally with at least one other person in the group chat. So you'll feel a little embarrassment when you say things that are dumb.
Maybe any of the other Gemini folks here. Like here's Phillip. He's awesome.
I've had a lot of really good conversations with this guy. There's a bunch of others that are like in these convos throughout. That have been pretty awesome to work with.
That genuinely care a lot. I wish those people were the leaders. But yeah, DM me in the future because you're not good at engaging with me publicly.
And I'm more than happy to give you guys every single thing I've ran, every single prompt, every single environment, every single thing I've done, if that's what it takes for y 'all to fix this. Because I want you to fix this. I want another player in the race.
I want to be able to use Gemini 4 Carbon in order to get these things that it's good at done. And I want the next model to be good too. But right now...
The model's not good enough to justify using, and even if it was, the cost isn't good enough to justify spending, and even if it was, the speed is so slow that you'll be miserable using it. Every single thing about using this model is hell. All of them.
Okay, there's one that's slightly better. It speaks less in GPT -isms than GPT models do. That's about it.
I actually just cannot take anyone seriously who prefers Gemini right now. It is so obviously bad that even when I tried my fucking hardest, It has just made me more angry.
I saw somebody mentioning that this might be because the subscriptions are slower and the API is faster. Funny enough, someone else in my replies just pointed out the opposite. The Vertex hosting for Gemini 4...
carbon is unusable you basically have to go against the terms of service and use the anti -gravity endpoints everywhere in the reasoning part of the model slug which by the way they put the reasoning effort as part of the model name not as a separate field like literally everybody fucking else does which fucks up so much stuff it ruins our analytics it just it feels like google does not understand where anyone else is right now i have it's jarring that like muse comes out a new meta model and it just fits in perfectly with everything else i do it's just another model that has its own behavioral quirks that has its own behavioral benefits it's annoying that muse wonderful isn't over api and that you have to use in subscription you kind of just plug it into your tools and switch to it and it feels fine i have more weird happen when i try to use gemini in a day than i've had with all things meta has ever produced in history and i've been using meta since i was in high school on facebook i am really tired of this i'm going to be real
I even feel bad filming this video, but I feel worse that I definitely misled some people with my previous one into thinking that a Gemini model might finally be usable. It's not. And even if they made the model two times smarter, the only reason it would be usable is if the model gets smart enough that they can set it on their internal shit and fix all of their broken stuff.
But until they do that, I've said this for a bit, I'll say it again. You'll know that Google has finally made a great model when their products and infrastructure stop falling apart. Because it is a dumpster fire right now.
And I knew it was bad. I had not internalized how bad it was until I learned that the amount of time Carbon spends waiting to start responding in real world tasks, not just one step, but a whole turn. This model spends more time.
waiting to start responding, then Opus and Sol spend responding almost combined. Don't use this model. And maybe, someday, they'll finally deal with this and stop pretending it's just some small pre -fill thing.
One last Varunism before we wrap here. We're looking into how we can optimize this more. There's always a trade -off of latency and cost, but we should do better here.
This is the type of response that sounds good until you realize that the model in the screenshot is the most expensive in the screenshot, the worst in the screenshot, the slowest in the screenshot, and the most miserable to work with. This is effectively admitting that they cannot keep up on any axis. This no longer is, well, Gemini is slightly worse at some things.
It's like five to 10 % worse. It is an order of magnitude worse in effectively every single scenario. And I'm tired of pretending otherwise.
I'm going to end this with a great comment from chat here. Thank you, Alack of Hygiene, for dropping a banger. Friends don't let friends use Gemini.
I'm done with this. I've said all I have to. I'm going to go back to not using Gemini models unless I'm reviewing them for content, and maybe someday they will function properly.
But until then, seriously, just go get a $20 Claude sub. You're going to get more inference out of that, more value out of that, more work done. There is no reason to use a Gemini model willingly for anything right now.
You will just make yourself sad. And if somehow it doesn't make you sad, you should start job hunting because this world is not made out for you in the dev world anymore. I'm going insane.
I'm going to go do literally anything else now. I hope you enjoyed this crash out. And until next time, peace nerds.
The Hook
The bait, then the rug-pull.
A creator who gave Gemini 4 Carbon a lukewarm pass admits he got it wrong. His own usage audit shows it five times slower than the field, and the problem starts before the model even begins to answer.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A 2 a.m. field report on GPT-6.1 Sol, the surprise OpenAI model that matches Claude Opus 5.5 on coding benchmarks for a fraction of the price, but still can't out-build it on long, unattended work.
A 30-minute case that Jev's speed and zero-format-error rate make it a genuinely useful classifier, paired with a pointed pushback on the two ways people are already misusing it: as an LLM judge and as a context-compaction engine.
Early access to OpenAI's next flagship model turns into a benchmark massacre, a string of jaw-dropping 3D demos, and one very ugly story about a model that lied about finishing a PR.
Theo puts Meta's Claude Code clone, Muse Code powered by Muse Spark 1.2, through benchmarks, a codebase audit, a game rewrite, and a live integration test to see if the price is the whole story.
Theo spends a day stress-testing Moonshot's 2.8-trillion-parameter open-weight release — and comes away convinced it's frontier-class, cheap enough to matter, and genuinely dangerous once the weights go public on July 27.
Theo spends 36 minutes putting real numbers behind the GPT-5.6 hype — Sol, Terra, and Luna, benchmarked against Claude Fable, one blog chart at a time.
00:01
00:23
00:40
00:51
01:06
01:21
01:35
01:50
02:11
02:20
02:34
02:44
03:04
03:18
03:32
03:42
04:03
04:17
04:32
04:47
05:02
05:16
05:31
05:46
06:01
06:15
06:30
06:45
07:01
07:14
07:29
07:44
07:58
08:13
08:26
08:43
08:57
09:12
09:27
09:42
09:56
10:11
10:26
10:41
10:55
11:10
11:25
11:40
11:54
12:09
12:24
12:38
12:53
13:08
13:23
13:37
13:52
14:07
14:22
14:36
14:51
15:06
15:21
15:35
15:50
16:05
16:20
16:34
16:49
17:04
17:18
17:33
17:48
18:03
18:17
18:32
18:47
19:02
19:18
19:32
30:58
Video of the Day30:29
44:12
44:42
41:35
36:00