Tools & Safety6 min read
What AI Does Commenti Use? Claude Opus 4.8, Explained
What AI does Commenti use? We draft LinkedIn comments with Claude Opus 4.8, Anthropic's most capable model. Here is why it matters, and what it will not do.
By Deepak Yadav, Founder, Ampliflow
The short version
- Commenti uses Claude Opus 4.8, Anthropic's most capable model, to draft comments in your voice on the posts you choose to target.
- Model quality matters for comments because a weak model produces generic, bot-sounding replies that get reported and ignored, while a strong one reads the post before it writes.
- Opus 4.8 drafts the comment, but it does not act alone: you can approve every comment before it posts, and it never sends connection requests or DMs.
- The model choice is about quality and safety judgment, not evading detection. LinkedIn's User Agreement prohibits third-party automation, including Commenti, and we say so plainly.
- A capable model also knows when to stay quiet: skip filters mean Opus 4.8 passes on posts like layoffs or grief instead of forcing a tone-deaf reply.
What the model does, and what it does not
It is worth being exact about the model's job, because "AI comment tool" tends to get imagined as more autonomous than it is.
What Opus 4.8 does: it drafts comments, in your voice, on the posts you choose to target. You point Commenti at the people, feeds, or topics you want to engage, and the model writes a reply for each one it is handed.
What it does not do:
- It does not post behind your back. You can run Commenti in approve mode and read every comment before it goes live. The model drafts, and you keep the last word. There is more on why that matters in our writeup on auto comments for LinkedIn.
- It does not send connection requests or DMs. Commenti is a commenting and engagement tool, full stop. It does not run outreach, it does not message strangers, and it does not scrape and blast. A tool that bundles all of that together is a different, and riskier, category of product.
- It does not decide your strategy. The model writes comments. It does not tell you who is worth engaging or grow a following on its own. That part is still yours, and a sound LinkedIn commenting strategy is what turns drafted comments into anything that adds up.
A better model is not a detection cheat code
This is the part most vendors skip, so we will put it in plain words. Using Anthropic's best model does not make automation allowed, and it does not make it invisible.
LinkedIn's User Agreement prohibits third-party automation, and that includes Commenti. There is no model good enough to change what the contract says. We picked Opus 4.8 for quality and for safety judgment, the ability to skip a post that should never get an automated reply, not to slip past anything. We do not claim Commenti hides from LinkedIn, and any tool that does is selling you a story.
Better-written comments do reduce one specific risk. Generic, templated comments are the ones humans report, and reports are a signal LinkedIn acts on. A comment that reads as genuinely yours is less likely to draw one. "Reduce" is the honest verb there, not remove and not eliminate. If you want the full picture of what actually raises and lowers account risk, we lay it out in is LinkedIn automation safe, and we apply the same lens to competitors in our roundup of the best comment automation tools.
How the model fits the rest of Commenti
The model is one piece. What keeps it useful and measured is the machinery built around it.
- Daily caps. Commenti's plans run at 20, 40, or 60 AI comments a day, priced at $19, $29, and $49 per month billed yearly. The caps are ceilings, not targets, because volume is the loudest automation signal there is.
- Human-like pacing. Comments go out spread across your active hours instead of in a burst, because nobody reads and replies on a metronome.
- Skip filters. The model passes on posts where an automated comment would land badly, like layoffs or personal news. This is where a nuanced model earns its cost, because it can tell the difference.
- Approve mode. You can review before anything publishes, which is the single biggest quality and safety control we ship.
Used this way, Commenti is built around roughly 20 to 60 new followers a day, though results vary with your profile, your niche, and who you engage. There is a 60-day free trial for our first 200 users, so you can watch the drafted comments in approve mode before you trust any of it to run on its own. The model quality is a big part of what makes that trust reasonable to ask for.
The straight answer
So, what AI does Commenti use? Claude Opus 4.8, Anthropic's most capable model, for the one job where model quality is the whole ballgame: writing a comment that sounds like you and fits the post. We use the top model because a comment is short, public, and attached to your name, and there is no room for the tells a cheaper model leaves behind.
Everything around it, the caps, the pacing, the skip filters, the approval step, exists so that quality is matched by restraint. And the honest footnote stays: LinkedIn's rules prohibit automation, ours included, and no model changes that. We would rather you know exactly what sits under the hood, and exactly what it will and will not do, before you decide.
Frequently asked questions
Which AI model does Commenti use?
Commenti uses Claude Opus 4.8, Anthropic's flagship and most capable model, to draft LinkedIn comments in your voice. We chose the top model on purpose, because writing a comment that sounds like you, and reading the post before replying, is a quality-sensitive job where a weaker model shows.
Does Commenti use ChatGPT?
No. The part of Commenti that writes your comments runs on Claude Opus 4.8, a model built by Anthropic, not by OpenAI. We keep the disclosure simple and honest: comment drafting runs on Anthropic's most capable model, because comment quality is exactly where a strong model earns its place.
Does the AI post comments automatically without my review?
Only if you let it. You can run Commenti in approve mode, where you read every drafted comment before it publishes, or let approved settings post within your daily cap. Either way, Commenti comments and engages. It does not send connection requests or direct messages on your behalf.
Does a better AI model stop Commenti from getting flagged?
No. A stronger model writes comments that read as genuinely yours, which reduces the risk from generic, reportable content. It does not change the fact that LinkedIn's User Agreement prohibits third-party automation, including Commenti. We chose Opus 4.8 for quality and judgment, not to hide the automation.
Why does a comment tool need Anthropic's most capable model?
Because the task is harder than it looks. A good comment has to read the nuance of a post, match your voice, avoid generic filler, and sometimes decline to comment at all. Opus 4.8 handles that judgment better than a lighter model, which is why we pay for the top tier.
Written by the team building Commenti on real LinkedIn growth data. Found an error? Tell us and we’ll fix it. Accuracy beats winning.
Founder, Ampliflow
Builds Commenti at Ampliflow, and has worked across 80+ startups as an AI micro-SaaS builder. Writes about what the product data actually shows, including the parts that are inconvenient. Commenti is the LinkedIn comment automation tool by Ampliflow.
What AI does Commenti use? Commenti uses Claude Opus 4.8, Anthropic's most capable model, to draft the comments it writes in your voice. That is the whole answer to the headline question, stated up front, and we would rather you get it in the first line than dig for it.
We build Commenti, a tool that drafts and posts LinkedIn comments for you, so we have an obvious reason to sound impressive here. This post is not that. It is the plain version: which model we use, why the choice matters for a job like commenting, and the parts the model does not touch.
Commenti drafts your comments with Claude Opus 4.8
To be specific: the part of Commenti that writes a comment runs on Claude Opus 4.8. Anthropic's current lineup includes the Claude 5 family, Sonnet 5 and Fable 5, plus Claude Haiku 4.5 for fast, lightweight work, and Opus 4.8 sits at the top as the flagship. We use the flagship on purpose.
Plenty of products reach for a smaller, cheaper model and accept the drop in quality, because for many tasks nobody notices. Drafting a comment in someone else's voice is not one of those tasks. The gap between the best model and a lighter one shows up in exactly the place a reader looks first: whether the comment sounds like a person who actually read the post, or like software filling a box.
Why a comment tool cares which model it runs
A comment looks like a small thing to generate. It is a few sentences. That is precisely why it is hard, because there is nowhere to hide. A weak model gives itself away faster in two sentences than in two pages.
Here is what the model is actually being asked to do, every single time:
A stronger model is better at each of these, and there is no shortcut around any of them. That is the whole reason we pay for the top tier instead of the cheap one.