Model capability is what harnesses expose, I have experimented with pi using local mlx models - several 27/35B quantisations (Gemma, Qwen, Flint) and the reality is that they are much weaker than Sonnet, let alone Opus. Of course they are "free" (if you have a $3000 mac) but they are slow and dumb, useful for very routine things but comparatively bad at problem solving and trouble shooting.

▲

aaronrobinson a day ago | parent | next [-]

I don’t think OP is proposing using a local model?

▲

throwa356262 a day ago | parent | prev | next [-]

Why do you limit yourself to mac?

You can run larger flash models on a modern PC using strata. These models perform pretty good even on low quants.

Furthermore, next generation open weight models will swap some compute for storage, making them more accessible for masses.

	▲	CarlitosHighway a day ago \| parent [-]
		Maybe he doesn't want to wear earplugs constantly and/or doesn't generate his own electricity?

▲

drTobiasFunke a day ago | parent | prev [-]

I am genuinely curious to know what kind of genuine problems people are solving with opus that cant be done with sonnet or the free models. If there are problems that people are working on that only opus 5.5 can solve then were we not touching such problems last year at all (because last years best model was inferior to sonnet 5.5). And if people are solving such problems now, why cant I see any meaningful improvement in even a single software I use everyday?

Not trying to diss on anything or anybody. I want to know.

	▲	magicalhippo a day ago \| parent \| next [-]
		Sonnet 5.5 feels like a cocky rookie. It's capable, but it often makes mistakes and especially poor assumptions, and when I point it out it just goes "yeah I did that wrong". Opus is much better at judging assumptions, makes less poor assumptions and mistakes. And when I point them out it takes it to heart and makes sure it guards against such a mistake in the future. Sonnet is great for simple work, such as subagents running a program and summarizing the output. Opus is just that much better at the more challenging work I find. You can probably get Sonnet to reach the same goal, but Opus needs much less hand-holding and overwatch.
	▲	aaronrobinson a day ago \| parent \| prev \| next [-]
		I took a few months out and it’s noticeable quite how much better opus is at least in respect of the quality of understanding. It’s a step change for me.
	▲	MINECRAFT_BIOLO a day ago \| parent \| prev [-]
		[dead]