@OtivDev profile picture
OtivDev @OtivDev

Working on 3D AI Modeling software - grandpacad.com

Joined May 2022
  • Tweets 208
  • Following 101
  • Followers 158
35 Photos and videos
"claude make me a SVG cheatsheet for this new tool I am trying out" - underrated use case
20
Opus 4.7 vs GPT 5.5 - 3D Modeling Benchmark (dozens of runs). About the same prompt adherence. But one of them costs 10x more and takes 3x longer. Guess which one, answer in thread.
1
96
Opus 4.7 sweeps GPT 5.5. No brainer which one I will be using now. @AnthropicAI
50
Thought Kimi K2.6 will replace Gemini 3.1 for 3D Model generation because of roughly equal intelligence & double the speed... Not yet. This screenshot is just one of a hundreds of comparisons I run. Kimi on the right.
1
66
which model should I try next?
23
Ran a bunch of 3D Modeling benchmarks on Gemini 3.1 vs Gemini 3. Unsurprisingly 3.1 performs a bit better. But surprisingly it costs 2.6x as much ($0.14 vs. $0.37 per 3D Model Generation) and is 2.5x slower (1m 24s vs. 3m 28s).
95
Reddit: "We hate AI slop in this sub." Me: "Here I created a sub for AI slop and volunteer to mod it". Reddit: "NO, downvote." Lol :D
63
OtivDev@OtivDev
12 Dec 2025
Ran 63 3D model generations with GPT-5.2. First it failed so much I had to update system prompt with a totally obvious instruction. Now with the updated system prompt it works but it's still slower, more expensive and generally worse than Gemini 3.
1
154
OtivDev@OtivDev
12 Dec 2025
It did however come up with a couple of solutions I never seen before.
1
43
OtivDev@OtivDev
12 Dec 2025
I am not even sure I want to run full extended benchmark for it.
36
OtivDev@OtivDev
11 Dec 2025
GPT-5.2 dropped. Tomorrow I will run 80 3D model generations to see how it performs. Curious to see if there are any quality/speed improvements vs Gemini 3.
1
2
120
OtivDev retweeted
OtivDev@OtivDev
19 Nov 2025
Burned through 18mil tokens for 3D Model generation benchmark to compare Gemini 3, GPT-5 and GPT-5.1. Gemini is clearly SOTA model. Findings in the thread.
1
2
8
382
OtivDev@OtivDev
19 Nov 2025
Burned through 18mil tokens for 3D Model generation benchmark to compare Gemini 3, GPT-5 and GPT-5.1. Gemini is clearly SOTA model. Findings in the thread.
1
2
8
382
more replies
OtivDev@OtivDev
19 Nov 2025
GPT-5 is steady but slow. Highest pass rate (80.95%) but averages 3 minutes per run. Reliable, yet too sluggish for this benchmark.
1
115
OtivDev@OtivDev
19 Nov 2025
Full blog post at: grandpacad.com/blog/comparin…
114
OtivDev@OtivDev
18 Nov 2025
Gemini 3 Released. Run 21 3D model generations. When it succeeds. It's much better. Not a hype man, just data. Bigger sample size later.
2
774
OtivDev@OtivDev
18 Nov 2025
I run over 200 3D model generations in last 24 hours. GPT-5.1 is about 2x faster than GPT-5 on the same medium reasoning settings. BUT it comes at a cost. It is less capable than GPT-5 and lands somewhere between Gemini 2.5 Pro and GPT-5.
1
167
OtivDev@OtivDev
21 Oct 2025
AI model I want: No reasoning (speed). GPT-5 level of performance. 20x faster. 4x cheaper.
68
Load more