Idk, they’re trying to sell a $500/mo/seat service to tell you what model is best. I think it’s in their interest to keep it confusing and opaque. Not exactly independent.
This benchmark gives the same intelligence score for GPT-6 Astra (max), GPT-5.6 Sol (max), and Grok 4.6 (high)? That seems very wrong to me, unless I'm misinterpreting the visualizations.
The most straightforward answer is that despite efforts to design a benchmark that, in theory, is supposed to measure generalizable intelligence, performance on ARC-AGI-3 can't be reliably correlated to performance anywhere else. I kind of lost faith in it after o1 or o3, I can't remember which, absolutely crushed ARC-AGI-1.
And, you know, maybe also some funny business. I think it's good to be a little suspicious of a model that happens to shoot upwards in performance on a specific benchmark while also kind of keeping up with the pack on a bunch of other benchmarks.
I don't think this is quite true. We have other examples.
Fable is without question the larger and more thoughtful/intelligent model. It also gets out performed by Opus on many/most benchmarks. So we can say that while Fable is more intelligent, Opus is more capable. I'd still opt for Fable in nearly every case if tokens were free.
So it can be true that the "smarter" model is perhaps not the smartest in every single niche dimension that its cousins have been fine-tuned for (yet!).
If you'd left them to the states, they'd have mostly been broken apart and sold off years ago, for starters. Especially when they see a budget shortfall.
This is a confusing situation then right? Taxation is the government using coercion to claim resources and labor. That's not to say it's evil, but the pool to claim resources from in a given year is fixed. So it's not clear to me it's a good thing that the federal government has a super power in that regard that states don't have. Anyway.. no particular point to be made here.
Most of the ocean is desert lacking the limiting nutrients of nitrogen, phosphorous and iron. If we can cheaply fertilize large swathes of the ocean with these limiting nutrients in a controlled manner, then this could trigger massive algae blooms that on sinking sequester massive amounts of carbon.
Would we have to continually fertilize the ocean? And at whose expense? It'd be great if such an operation could be profitable, but it there's only so much seaweed people can eat.
Many crop fertilisers have that issue, but fertiliser is not only one thing.
One hypothesis is that algae are limited by some single resource which could be gradually dropped out the back of cargo ships as they cross oceans, seeding carbon-absorbing algal blooms as they go.
(I heard about this nearly 20 years ago, so I assume that has either been tested or banned since then…)
Indeed, that's one of the main possibilities I guessed you were thinking of; for triggering algal blooms, one of the things suggested was iron. (And given it was immediately going into salt water, ore would presumably have been fine… if the hypothesis was correct).
Terrestrial ag fertilizers don't have to be produced by fossil fuels.
The Haber–Bosch process used to make ammonium nitrate can be solar powered instead of using methane.
Phosphate rock processing could greenified for a green "green" phosphoric acid wet process.
Potassium from potash is mostly a passive solar evap process.
For seaweed aquaculture, iron sulfate is used because iron availability is the most frequent limiting factor. It's frequently a byproduct of steel production, in much need of greening itself, and produced in large quantities. Volcanic ash is another abundant source.
it depends on the final balance. As mentioned above the ratio for iron is 13k Carbon atoms sequestered for atom of iron.
Keep in mind the vast majority of agricultural yield growth the last century is giving these limiting nutrients as fertilizer for plants, so it is successful and worthwhile for land based farming atleast.
There are some similarities, but also some differences. Flower's take is that it wants to support the entire FL workflow from experimental research to large-scale production deployments and operation. Some other FL frameworks fall either in the "research" or "production deployment" bucket, but few have good support for both.
Flower does a lot under the hood to support these different usage scenarios: it has both a networked engine (gRPC, experimental support for REST, and the possibility to "bring your own communication stack") and a simulation engine to support both real deployment on edge devices/server and simulation of large-scale federations on single machines or compute clusters.
This is - to the best of our knowledge - one of the drivers of our large and active community. The community is very collaborative and there are many downstream projects in the ecosystem that build on top of Flower (GitHub lists 748 dependent projects: https://github.com/adap/flower/network/dependents).
https://artificialanalysis.ai/models
Perhaps if it was allowed this custom harness for all benchmarks it would similarily saturate?