Main announcement · Blog Now that Grok 4.5 is available, we’d love to hear how it’s working for you. Some things we’re especially curious about: What are you using it for beyond coding? This is the first model we’ve built for more than software engineering. If you’ve pointed it at data science, research, finance, legal work, or anything else you do on a computer, we want to hear how it went. How does it handle long-running, multistep tasks? Grok 4.5 was trained to investigate problems, use tools creatively, recover from mistakes, and verify results before calling it done. Does that match your experience? How does it compare to Composer and other frontier models in your day-to-day? Anything noticeably better (or noticeably worse)? It should be solving tasks in notably fewer steps; let us know if you’re seeing that. This is also our first model trained jointly with SpaceXAI, and the kind of multistep, tool-heavy work it’s built for is exactly what’s hardest to capture in a benchmark, so your feedback shapes where the model goes next. Drop your thoughts below (and leave a Request ID if you have one)! We read everything. :folded_hands:

[image] Someguy: @Colin the double included usage that’s like always from now on? Not a promo? And how is this done during the current month percentage counts, since I didn’t see it becoming less percentage in my spending? It’s not a promo, and the increased usage already applies to this month!

In panning mode Grok 4.5 doesn’t trigger then proper question UI and just asks via text

I’d like to try out 4.5, but I have a degree of reticence in using it. Given that Cursor openly states that it was trained using user data (Cursor · Grok 4.5), is there a hard and fast commitment on Cursor / SpaceXAI’s part to honor customer privacy preferences with respect to past, ongoing and future use of data for model training, especially when Privacy (Legacy Mode) is selected? For instance, https://apidog.com/blog/grok-4-5-cursor-training-data/ raises some questions about this. I’m a Cursor fan, and presume good intent, but haven’t seen these questions answered definitively (Cursor’s privacy policy is actually quite clear, but I’m thinking more about the commingling of privacy policies via the Space(XAI) acquisition, and whether these commitments will continue now that Cursor has been acquired, and where the four corners of “Inputs or Suggestions” (Cursor’s term) apply or don’t apply to data use and how this crosses Cursor and xAI API processing boundaries. As the founder of an AI governance platform, I’m especially focused on understanding how our intellectual property is being handled and how individual models process customer data.

Grok 4.5 just spun up Opus 4.6 subagents. Completely ignoring that i selected a first party model. Meaning the billing switched over from the first party models to the API models. Which I had already maxed out, luckily i had set a max on on-demand. Because it completely went rogue and used ALL my usage. This is absolutely terrible. Even though i selected a first party model, which should pull from the first party models usage, it just ignored it and blazed through my API. Thanks for the bill Grok… :face_with_steam_from_nose:

I’m not a founder, so I don’t think the same way you do. But after seeing what Cursor could offer me for the price, I actually turned Privacy off, hoping that my workflows would contribute to making future versions of Grok and Composer as good as possible.

[image] Remi Hindriks: Even though i selected a first party model, which should pull from the first party models usage, it just ignored it and blazed through my API. Just set up Subagent cards, specifying which model to use and when. Barring a few bugs, Grok works perfectly with them.

It worked until few minutes ago. Now I get: “This model does not support custom API keys.” What is going on? PS: I disabled the local models I was playing around with few months ago and it worked. But that’s just a bug. I never had problems since I added them until now.

Hah. Had a very similar experience a few months ago – Better subagent control in the Cursor IDE . @Colin 's advice at the time helped me, as our fellow poster pointed out – a subagent rule is effective in governing which models get invoked for this purpose (Composer 2.5 has worked really well for me here).

I see this from time to time after the model burns a lot of tokens and time (I though it crashed): "The large replacement failed. I’ll use smaller patches or a targeted write instead. " Did the harness just block a large patch ?

[image] rsb: is there a hard and fast commitment on Cursor / SpaceXAI’s part to honor customer privacy preferences with respect to past, ongoing and future use of data for model training, especially when Privacy (Legacy Mode) is selected? Cursor Grok 4.5 is subject to the same Privacy Mode controls as other models offered in Cursor. With Privacy Mode enabled, Customer Data is not used for training by Cursor or model providers, and Cursor maintains zero data retention agreements with model providers for ZDR-supported models.

Share your Thoughts on Grok 4.5!

Announcements Release Discussions

Artemonim (Artemonim) July 18, 2026, 12:48pm 150

Just set up Subagent cards, specifying which model to use and when. Barring a few bugs, Grok works perfectly with them.

Topic		Replies	Views
Grok 4.5 Removed? Help grok	38	1418	September 10, 2026
Share your Thoughts on Grok 4.6 Release Discussions grok	87	6568	November 9, 2026
Share your Thoughts on Grok 4.7 Release Discussions grok , composer	136	8986	November 5, 2026
Grok free on Cursor - Feedback needed Discussions	71	6451	September 24, 2025
Cursor IDE force enabling Grok 4.5 and setting as default model, regardless of my initial selection Bug Reports grok	16	1523	September 29, 2026

Share your Thoughts on Grok 4.5!

Related topics