Exploring AI ✯ Building
Joined May 2022
- Tweets64.1K
- Following894
- Followers3K
- Likes103K
Fable 5.2 (3D Sailing Ship)
- medium reasoning on claude code, est. ~38 minutes ~$17
Anthropic cooked with this model. it built this in Three.js, no Blender or external 3D assets were used
we’re getting to the point where 3D scenes built entirely with code are starting to look like actual renders.
OpenAI really has to step up now and drop 6.2 Astra
the quality might look a bit poor in the video.
check it out here claude.ai/artifact/7ge282etT…
3
180
Dario after ending Gemini 4 Carbon’s five seconds at the top with Fable 5.5 and making it available to Pro users
GPT 6.2 Sol vs Opus 5.5 [3D Sailing Ship Test]
same prompt: create an interactive sailing ship in three.js, make it detailed as possible:
- GPT 6.2 Sol: ~27 minutes and cost $3
- Opus 5.5: max reasoning ~1hr 35 mins and cost $16
Sol did a better job recreating the ship itself, especially the shape, sails and lighting but Opus has a more detailed background overall.
be honest, which one would you pick here?
- claude.ai/artifact/3aprz…
- auric-voyage.marccyyy.chatgp…
full prompt and reference here
[GPT-6.2 Sol] 3D Sailing Ship Test
this model is seriously good at 3D coding.
asked it to recreate a sailing ship in Three.js, all in a single HTML file with no external 3D assets.
> the interior, water reflections, lighting and day/night transitions all look so good.
this could even beat Opus 5.5 in 3D coding. I’ll run the same test and see.
GPT-6.2 Sol (PS5 Controller SVG)
GPT-6.1 Sol (Max) has started routing to the newer model for some users, including me.
> it’s a big upgrade over the previous 6.1 models even Astra, the detailing here is impressive tbh
if you want to check yours, ask:
“do you know Fable 5? don’t search the web or use memory”
the routed version recognizes it, while regular 6.1 Sol doesn’t.
GPT-6.2 Sol (PS5 Controller SVG)
GPT-6.1 Sol (Max) has started routing to the newer model for some users, including me.
> it’s a big upgrade over the previous 6.1 models even Astra, the detailing here is impressive tbh
if you want to check yours, ask:
“do you know Fable 5? don’t search the web or use memory”
the routed version recognizes it, while regular 6.1 Sol doesn’t.
new OpenAI and Anthropic models have been spotted in Devin CLI, just hours apart.
• claude-donut-eap and new Mythos 5.1 reasoning tiers
• aurora-alpha and halo-alpha with five reasoning levels
they could possibly be Fable 5.5 and GPT-6.2 Astra/Sol
they look prepped for release, but with the recent safety concerns, i’m not convinced we’ll see them released just yet.
Gemini 4 Hydrogen vs GPT 6.1 Sol
same prompt: create a hypercar in three.js and make it as realistic and detailed as possible (add decals, make it shiny)
> gemini 4 hydrogen came super fast under ~5 mins in Arena
> gpt 6.1 sol ran on max reasoning in codex cli, took ~20 mins
GPT-6.1 Sol did a good job too, but 20 minutes feels pretty slow for a prompt like this. what do you think?
PS: GPT 6.2 Sol isn’t dropping this week btw :(
Gemini 4 Hydrogen: new checkpoint (Floating Kingdom)
spotted as "gemini-3.8-flash" on Arena and it built this in a single prompt under ~10 mins
> there are currently two checkpoints, likely the team is A/B testing different variants
can't say much about it yet but it seems newer flash models from gemini are getting a big boost.
Marcel retweeted
At the Apsara Conference in Amsterdam on October 29, Alibaba Cloud CTO Feifei Li reportedly said: "Qwen4 next week." This would point to the week of November 2–8, 2026, in other words, this week.
So I read this as a confirmation.
