Not exactly what you asked for but try https://pelicans.borg.games/

This is also interesting: https://static.simonwillison.net/static/2025/pov-pelican-opu...

AFAIK that model is pretty old, and it was explicitly trained for SVG generation. For other models the capability of generating SVGs of real stuff is accidental. Same as GPT-5.x and Sonnet 4.5+ being able to generate MIDI music.

is it a fine tune of some open source model?

In US downloading copyrighted data is not illegal AFAIK

How many years did it take for you to write a Photoshop clone eating Adobe's lunch?

Couldn't tell if the first part was a sarcasm or not.

The second part gave it away though.

Tuned Qwen 3.5 27B beats Step 3.5 on almost all benchmarks, so the point about the size class is moot.

Benchmarks are not interesting in deciding the "size class". Bigger size means more knowledge. Also, the Qwen 3.5 27B is a dense 27B active parameter model. StepFun 3.5 Flash has 11B active parameters.

> Bigger size means more knowledge.

Qwen 3.5 27B beats StepFun 3.5 Flash on GPQA Diamond too, so probably no.

Benchmarks don't tell the whole story. For one-shot coding tasks, I found Step 3.5 Flash to be stronger even than Qwen 3.5 397B.

Benchmarks don't tell the whole story... for that you need anecdotes from random HN posters :)

I would argue that flattery without calculation is just poorly calculated flattery.

Same applies to many other traits in the list. Low achievements people lie right and left just as well. Are cynical when convenient, yada yada.

Basically, the list says that these 30s are just like an average Joe, but smart. Which should be a surprise to no one.

No, because smart people realize they are playing an iterated game and that behaving in a way that people identify as Machiavellian is actually suboptimal in the long run.

So they're smart enough to be calculated and stupid enough not to be so calculated that they look untrustworthy.

> No, because smart people realize they are playing an iterated game and that behaving in a way that people identify as Machiavellian is actually suboptimal in the long run.

Even if you are right coincidentally (which I wouldn't be so sure about), that's still poor argument assuming you realize your belief in what optimal strategy is what it is - just an educated guess.

Why does that matter?

They're only the same thing if you ignore intent.

Not everyone lies or is cynical when convenient. Skill, rate of success, and personal ethics are all orthogonal concepts.

Above all, intent matters. I do not treat someone who I perceive to be manipulative the same as I would other people.

> all orthogonal concepts

That's my point exactly. I just also assume the % of liars in 30s is the same as in general population by default.

In large providers KV caches are the main bottleneck, no?

Last time I checked Codex didn't have that option for $20 plan

I also really liked Foundation on Apple TV