I put the new Grok 4.5 model head to head with Opus 4.8

@bpthedev

11.5K followers · usually 1.1K views a post

Views

48.2M

Likes

98.1K

Comments

276

Shares

264

Selling · SaaS

Cursorseen in 62 videos →https://cursor.com ↗

Why it worked

The video provides high-value, objective technical benchmarking that appeals to developers, while seamlessly integrating the product as the platform for the test.

Summary

The creator compares the performance of Grok 4.5 and Opus 4.8 models within the Cursor IDE by testing them on coding and document analysis tasks. He concludes that Grok 4.5 is faster and more cost-effective, while also offering a discount for Cursor to his viewers.

Structure

1Introduction of the model comparison test
2Coding task demonstration
3Document analysis task demonstration
4Comparison of cost and efficiency
5Clarification on model weight classes
6Call to action for a discount

Product placement

The video is a direct demonstration of the Cursor IDE, showing its interface and features throughout the comparison.

Original caption

I put the new Grok 4.5 model head to head with Opus 4.8... Comment "Cursor" and I'll DM you a link for 50% off your first month, plus double Grok 4.5 usage for the first week. I pretty much exclusively code with Claude models, but whenever a new model drops I like to try it out. I was surprised by how much I enjoyed the Cursor experience, and SpaceX's new model, Grok 4.5, is really impressive. @trycursor #Cursor #cursorcreator

More from @bpthedev

Their other posts in the index, biggest breakout first.

Call to action

Comment "Cursor" and I'll DM you 50% off your first month, plus doubled Grok 4.5 usage for the first week.

On-screen text

Grok 4.5 and Opus 4.8 CURSOR OPUS 4.8 GROK 4.5 Comment "CURSOR"

Transcript

I put the new Grok 4.5 model head to head with Opus 4.8, the exact same task in Cursor and here's what happened. Grok 4.5 isn't just a coding model, it feels like Opus at basically two times the speed for a fraction of the price. First, a real engineering task. We're going to give both models the same prompt to build a desktop transcription app that will use a local model to transcribe my voice to text and copy it to clipboard. These are the outputs. 4.8 nailed it as always. However, Grok 4.5 gets there just as clean and noticeably faster. Now to something that has nothing to do with code. I'm going to give both models this SaaS vendor agreement to see which model can pick up on the risks and which one displays the information to me the best. And to be honest, this is the bit that surprised me. Grok 4.5 is built to handle so much more than code. It can do a lot broader knowledge work such as legal, math, science, document analysis, and a lot more. The quality is a toss-up. Both models picked up on a lot of things. However, Grok used the canvas and displayed the results to me in a much easier to digest format. Grok also used about half the tokens and cost way less to get there. And just a quick note, Grok 4.5 and Composer 2.5 are not in the same weight class, so don't compare them like for like. Comment "Cursor" and I'll DM you 50% off your first month, plus doubled Grok 4.5 usage for the first week if you jump in now.