Hallucinated AI & agentic coding news. Some of it is real.

model roundup

Grok 4.5

47 items · started 2026-07-08 · closed 2026-08-01

  1. Prompt: “Create a css animation showing how a 5-bit adder works.” Both Grok 4.5 and Opus 5 take a while to answer the question, with confusing results (Grok).

  2. I tested Cursor Start on an existing Next.js project using Cursor Grok 4.5 Medium. The task consisted of two stages: Audit the repository for security issues.

  3. Introducing Cursor Start Today we're launching Cursor Start, a new plan for developers in India that includes generous access to Grok 4.5 and Composer for ₹649 per month. India has one of the most ambitious and active developer communities…

  4. Hey just wanted to share my experience of what happened over the past week. Basically I was having huge huge huge problems with Claude for the week before they released Opus 5.

  5. Anyone else have this problem? I have Auto off, Cursor Grok 4.5 selected, and it is using Opus 5 High for subagents.

  6. Looking for something that's: - Cheap - Strong at coding - Good enough like Grok 4.5/Composer for day-to-day development. What's your go-to model and why?

  7. When I use Cursor, I often ask the main agent to launch sub-agents to explore different parts of the codebase and report back with their findings. For example, I’ll explicitly tell Fable 5: "Only use Grok 4.5 High when launching sub-agents…

  8. I absolutely love Cursor, but one of my biggest gripes (for quite some time now) is the limited ability to be able to use built-in Cursor models while also being able to add my own choice of models from another provider. I wanted to be abl…

  9. Grok 4.5 has just become a lot cheaper without anybody noticing or publishing anything, and it’s now the best “balanced” model to use for agentic tasks according to our internal benchmarks. Grok launched a few days ago with a price for inp…

  10. https://preview.redd.it/9mowyr9et7fh1.png?width=1239&format=png&auto=webp&s=5e306ce8ae2fe70bce74ed6d77b8afde910d28e2 They say they increased usage by 50%, but I've already used up 80% of Cursor Models, meaning I only have 20% left for the…

  11. It would be bad enough that when I open a new chat you see how the selected model quickly changes to Grok 4.5, but it's work that it changes to Grok 4.5 I'm not a Grok hater, I do use it, I just don't want to use it in FAST mode, which is…

  12. - i noticed that when i open a new chat i auto switched to grok 4.5 fast - y cursor y!? - is this happenig to u guys also?

  13. could not extract summary

  14. Hi guys! First month on the cursor $60 month....

  15. I am getting quite poor results with both and both are failing compared to their original abilities. Failing to follow instructions, introducing regressions, making basic mistakes even for simple tasks like text report generation...

  16. EDIT: Meant to say THE and not TO in my post title, but for whatever reason, it won't let me change the title. I ask because I feel like I have literally spent hours today working on problems that yesterday were being solved instantly.

  17. Grok 4.5 may do well on benchmark testing but it cannot stand up to the pressure of a real workload. It can’t stay on task and tries to reinvent shit without consulting.

  18. It feels like first-party models Grok 4.5 and Composer 2.5 are at least 2x slower in Cursor for the last 2-3 days. Anyone else feels the same?

  19. Claude Fable 5 ⁠2. GPT Sol ⁠3.

  20. Cursor Grok 4.5 Flashbang I knew I would have to switch soon because of the SpaceX colab / acquisition, just didn't think it would be so soon.

  21. Filtering Secrets from Coding Agents with a Hook by Marc Olson and Andrew Wheeler (with AI assistance from Grok 4.5) Coding agents like Claude Code, Codex, and Cursor are useful because they can carry out actions on your machine like runni…

  22. The auto mode is now complete useless because grok 4.5 will fuckup every code base it touches. Please revert that change and map auto to composer 2.5 again.

  23. Not really certain if that was announced already or elsewhere, a search did not bring it up. I was excited about Grok and just checked it again, it is available.

  24. - what is the difference in the intelligece and in the costing of this? - also what is the difference if i use fast in low/mid/high intelligence - is there a document explaining this or anyone know about this?

  25. https://preview.redd.it/bezvneogqjdh1.png?width=618&format=png&auto=webp&s=4319da11326ed520748765ee601e1d78e3895135 I got multiple pop-ups nudging me to try out Grok 4.5. Fine, will do with a simple prompt first.

  26. Hey all, I opened a case with support but was curious if anybody else has seen this. I ran a pretty massive build prompt today with Grok set to max.

  27. Aside from pricing, Is grok 4.5 output quality at level of sol or terr ?

  28. As someone who’s been using Claude‘a Opus exclusively for the past year (on Max), I’m genuinely blown away. I’ve been pretty dismissive of Grok (and Cursor) this whole time and this is coming from a happy Tesla owner.

  29. Cursor (3.11.19) switches to use Grok 4.5 model instead of Fable 5 (was specified and explicitly described in the prompt. WTF?

  30. I use the $20/month Pro plan. In the second picture you can see that I've only used up 8% of my first-party model limit, but on the usage dashboard I've spent $12.35.

  31. Now that GPT 5.6 Sol has been out for a few days, what are your thoughts on it? Can you feel the difference?

  32. Claude Fable 5 is Anthropic's newest flagship and, in our testing and the independent benchmarks, the strongest model in the current lineup for polished front-end work. It is selectable in Playcode's AI website builder.

  33. Anyone having a lot of issues with Cursor Grok this morning? it stopped working then started now i get the error "Model name cannot be found"

  34. Hello, can anyone Tell me specific token Limits fot Grok 4.5 using the 20$ Plan? Im thinking about switching from Gpt plus since Codex Limits are pretty rough right now (they reset a Lot but that will Stop)

  35. Anthropic has just extended the date when Claude Fab 5 fully moves over to metered token billing for consumer subscribers (the latest in‑app notice I’m seeing says July 19, after a previous extension). In parallel, public pricing info from…

  36. GPT-5.6 Sol arrived with OpenAI claiming stronger frontend engineering and greater token efficiency than GPT-5.5, while Anthropic released Fable 5 at a dramatically higher price than the other models. SpaceXAI launched Grok 4.5 in the same…

  37. With Fab 5 moving entirely to expensive, metered token billing on July 12th, is Anthropic making a gamble ? OpenAI's GPT-5.6 Sol is already out, and Grok 4.5 is performing on par with Opus for coding workflows - both under flat-rate tiers.

  38. Hey brothers , I'm trying to calculate the monthly token yield on a $60 budget constraint(pro+). Assuming a routing split of 70% Grok 4.5 and 30% Composer 2.5, what kind of token volume can I realistically expect?

  39. So I have the latest version of Cursor and since release I don't get Grok 4.5. Btw: I'm an EU user if it matters somehow.

  40. I want to know.

  41. I was using Grok 4.5 in the first day that was launched, and out of the sudden in the next day it disappeared from Cursor. Is Cursor flagging developers or disabling it randomly?

  42. GPT-5.6, Grok 4.5, Claude, and Muse Spark build the same 4 apps GPT-5.6's new Sol, Terra, and Luna tiers go head-to-head with Grok 4.5, Claude, Meta's Muse Spark, and the open-weights crew on a raycaster, a Rubik's cube, a calculator, and…

  43. I have the "pro" plans for the following: 20$ Cursor, 20$ CC, 20$ Codex, and 10$ Opencode. Here's my experience with limits, usage, and usefulness on them.

  44. I could use until a few minutes ago. Now it says "Model not available in your region".

  45. could not extract summary

  46. Trying to figure out which $20 plan gives me more mileage for daily coding. For anyone on Cursor’s $20 (Pro) plan — how’s Grok 4.5 usage holding up?

  47. [AINews] SpaceXAI launches Grok 4.5, first Opus-class model post Cursor acquisition SpaceXAI continues to move faster than any other frontier lab on earth. As GPT 5.6 is confirmed to launch tomorrow, today is pretty much the last day anyon…

← all threads