CrossstarDesigns

Blender 3D Artist & Addon Creator | Developing AI tools for 3D workflows | Product Visualization | Commissions Open

[CrossstarDesigns @crossstardesign]

@diegocabezas01 yeah this one feels like it cheated. sonnet actually tried and followed the prompt. 11k paths is just a trace with extra steps

[CrossstarDesigns @crossstardesign]

@melvynx Opus 5.5 medium is the most balanced model right now smart enough without wasting tokens like the bigger runs

[CrossstarDesigns @crossstardesign]

Same prompt on Opus 5.5

[CrossstarDesigns @crossstardesign]

Sonnet 5.5 Medium looks better than GPT-6.1 Sol XHigh. And you can actually use it on the $20 plan.

Sol only starts looking decent at XHigh. That’s the setting that eats the limit.

[CrossstarDesigns @crossstardesign]

SuperGrok's grok 4.5 usage limits are worse than cursor. In cursor it's included with a generous auto quota but here you burn through it in just 2 days. Why pay an extra $10/month for that? If you're a developer just choose Cursor Pro or pro+ until @elonmusk @SpaceXAI increase the limits.

[CrossstarDesigns @crossstardesign]

@RiyaDesaihere Thanks!

[CrossstarDesigns @crossstardesign]

Jewelry CGI where a ring drops straight from the sky onto the Eiffel Tower. Tracked the real footage using Syntheyes, animated everything in Blender, rendered with Octane, and did the post + rotoscope in After Effects and sound design done in Ableton.

#b3d #blender

[Chris Laupama @chrislaupama]

I used the @cursor_ai $20 Pro plan and found that using GPT-5.5 for planning and then Composer 2.5 (non-fast is 6x cheaper than fast) for building was the perfect balance and got me incredibly far.

It’s the plan I’m going to recommend to my friends and those wanting to build with a mix of power and efficiency.

The bonus is you can use any other model (Gemini, Opus etc) if you want at any time too!

[CrossstarDesigns @crossstardesign]

@wxrrjxr @LottoLabs @aiCrof @ollama @ollama needs to update the plan. 50x usage shouldn't mean hidden token limits. Seems they quietly shifted from gpu time to tokens, scanning code instantly eats 10-15% of the 5hr limit, unlike regular text. Feels like they're trying to cover new Blackwell GPU costs.

[CrossstarDesigns @crossstardesign]

@wxrrjxr @LottoLabs @aiCrof @ollama Looks like @ollama drastically cut the limits. A 5 hr limit getting wiped out in minutes on just a couple of tasks now and the weekly limit is heavily reduced as well.

[CrossstarDesigns @crossstardesign]

Same task, same project:

@opencode Go + Kimi K2.5: ~200 requests = only 4% monthly quota
@cursor_ai Pro + Composer 2: ~60 requests = 15% generous pool

Composer 2 uses 13–30× more tokens per request (1M–2M vs ~70K).

Huge efficiency difference.

[CrossstarDesigns @crossstardesign]

@cursor_ai Yesterday, after just 1 hour with only Composer 2, used 3% generous Auto + Composer and 6.5% API. I didn’t use Cursor anymore after that. Today, when I opened the app, it suddenly jumped to 8% generous and 16% API with zero additional work.

[CrossstarDesigns @crossstardesign]

@cursor_ai After just 1 hour with only Composer 2 (41 requests, no other models), I’ve used 8% of the generous Auto + Composer pool on the pro plan.

I stay exclusively on Composer 2 Standard and have optimized everything. Also getting Provider Error

[CrossstarDesigns @crossstardesign]

@aye_aye_kaplan It would be great if Cursor offered a more generous Composer 2 focused Pro plan for users who only use Composer and don’t really use API credits. That kind of option would be incredibly valuable

[CrossstarDesigns @crossstardesign]

@cursor_ai would be amazing if Cursor offered a more generous Composer 2 focused Pro plan for users who mostly rely on Composer and don’t really use API credits. That kind of option would be incredibly valuable for long term builders

[CrossstarDesigns @crossstardesign]

@SkylerMiao7 M2.1 was already a massive step up. If M2.5 is even crazier, I’m hyped. Will we finally see multimodal (images/video) and browsing support?

[CrossstarDesigns @crossstardesign]

@opencode Cached input: $0.08/M (opencode) vs $0.10/M (direct) so only $0.02 savings
Regular input: same $0.60/M
Output: same $3.00/M

So the 20% cut applies only to cached tokens? The “cheaper than Sonnet/Opus” headline suggests much larger savings. Am I missing something?

[CrossstarDesigns @crossstardesign]

I got tired of converting Cycles materials to Octane manually. So I built an addon that does it automatically.
Watch the demo.

Releasing later this week. Would this save you time?

#b3d #Blender3D #OctaneRender #BlenderAddon

[CrossstarDesigns @crossstardesign]

@imaginationIRL Do you just need 3D modeling or Visualization too?