Honestly, the numbers are becoming increasingly difficult to interpret. Every time a new version comes out, they just call it the "best." It would be much more useful to directly compare performance on sets that people actually use, such as coding and summarizing.
Gemini 3.5 Flash
531–540 of 692 posts
Re: Gemini 3.5 Flash
#532Earlier quoted context omitted.
It was probably the right call at the time with low bandwidth. Nowadays I bet flash would execute faster than most js heavy sites :D
It was not the right call, Steve Jobs was just a monopolist killing a competing platform and we're all worse off for it.
Flash was a security and battery disaster just like Java applets. Both are dead as browser plugins and good riddance.
Re: Gemini 3.5 Flash
#533Am I really so old that when someone says "Flash" my immediate response is... "consider HTML5 instead" ??
Frontpage, Dreamviewer, flash, photoshop lol. We are old.
Re: Gemini 3.5 Flash
#534For those who would like to know the total and active parameter count of this model: even though Google doesn't disclose the model technicals, we can infer them within relatively tight margins based on what we do know. We know they serve the model on TPU 8i, which we have plenty of hard specs for (so we know the key constraints: total memory and bandwidth and compute flops). We can also set a ceiling on the compute c…
We've been really impressed with the performance of ~30B parameter class models and how close they are to the frontier from ~6-12 months ago, which begs the question, are the frontier labs really serving 10T parameter models? Seems unlikely. If these Gemini 3.5 numbers are accurate, then I'd wager GPT 5.5 and Opus 4.7 are a lot smaller than people have speculated, too. It's not that frontier labs can't create a 5T+ p…
I agree with this sentiment but the reasoned anecdotes do not agree. I imagine the flagship models have modalities/usages that we hn-ers don't imagine easily.
Re: Gemini 3.5 Flash
#535Earlier quoted context omitted.
All major operating systems Windows, macOS, iOS, and Android have local APIs for using AI.
Why would I use those instead of just grabbing a model from hugging face? Are they as good as qwen 30B?
Because it is simpler as an application developer to just use an OS API then trying to figure out some 3rd party thing and setting that up. Each platform has several different models for different things so I can't give a comparison.
Re: Gemini 3.5 Flash
#536Per million input/output tokens: Gemini 2.5 flash: $0.30/$2.50 Gemini 3.0 flash preview: $0.50/$3.00 Gemini 3.5 flash: $1.50/$9.00 Interesting pricing direction. I don't think we have ever seen a 3x price increase for in the immediate next same-sized model (and lol @ 3 only ever getting a preview). 3.5 flash costs similar to Gemini 2.5 pro which was $1.25/$10
This understates the cost increase. 3.5 Flash also uses more tokens. artificialanalysis.ai shows these difference to run the whole eval, which I think is more realistic pricing: Gemini 2.5 flash (27 score): $172 (1.0x) Gemini 2.5 pro (35 score): $649 (3.8x) Gemini 3.0 Flash (46 score): $278 (1.6x) Gemini 3.5 Flash (55 score): $1,552 (9.0x or 2.4x compared to 2.5 pro) This is a massive price increase... 5.6x compared…
the era of subsidised ai is ending
Re: Gemini 3.5 Flash
#537Earlier quoted context omitted.
Amazon was unprofitable for over a decade, and they were public. Theres no incentive to be profitable as a private company if you can continue to raise money. Ed Zitron and Gary Marcus are... confused.
> Amazon was unprofitable for over a decade, and they were public. Amazon was unprofitable because they poured their revenue into growth. On paper, they were in the red, but everyone - especially investors - saw what was going to happen, given their trajectory. Is it the case that any of these AI companies are actually making a ton of money and growing accordingly? AFAICT, we've just got [a] big players like Google t…
> On paper, they were in the red, but everyone - especially investors - saw what was going to happen, given their trajectory.
As I recall, no, Wall Street and public shareholders were getting pretty antsy over AMZN earnings, which is why Bezos famously said "We are willing to be misunderstood for long periods of time."
The same thing is playing out today: insiders and early investors (presumably privy to information we don't have) see the trajectory of the frontier AI labs, but Wall Street and public shareholders see only the losses. This is why at every earnings report the hyperscalers simultaneously 1) post record revenues and earnings, 2) announce even greater CapEx spending and AI investments, and hence 3) get punished by the stock market.
Clearly all the AI players are willing to be misunderstood for long periods of time.
Re: Gemini 3.5 Flash
#538What happened to gemini 3.2, 3.3, and 3.4..?
Re: Gemini 3.5 Flash
#539Earlier quoted context omitted.
Rumor is that GCP was happily selling compute to competitors. After all, under the hood, Google is closer to a federation than a corporation. The state of GCP doesn't care about the state of Gemini.
> Rumor is It’s not a rumor - there are many public announcements about $B deals around compute for other Ai companies
>> Rumor is that GCP was happily selling compute to competitors. After all, under the hood, Google is closer to a federation than a corporation. The state of GCP doesn't care about the state of Gemini.
> It’s not a rumor - there are many public announcements about $B deals around compute for other Ai companies
The last time I read a public announcement, the commentary I read was this is because Anthorpic doesn't want to run out of cash or capacity before a funding round / IPO so they gave Google some equity and in return Google gave it some compute resources? You could reframe it as Google is buying into Anthorpic — which is how Claude tends to frame it as but the end result is the same. Equity for spare capacity.
You could even argue that at a hyperscaler like Google's scale — all capacity is spare capacity and no capacity is spare capacity at the same time. GCP seems to have deals with Anthorpic, OpenAI, Meta, Apple, Healthcare companies, Banks, LG(?), Best Buy(?) so in my mind Google (and all AI vendors) are hyping up AI to drive up interest and building up as fast as they can to capture that interest and convert it into cold, hard cash. It honestly feels like this is out of my mental capacity though because these AI vendors had the cold hard cash that they spent on these data centers that we don't even know might become obsolete within a decade(?) but I guess meanwhile they could make beaucoup bucks. There is also the idea that they had to be seen as conspicuously spending on AI or investors might see them as falling behind, triggering a selloff. So yeah I guess it is a fact that GCP is selling compute to other AI companies but it makes sense because basically you can build capacity potentially for Gemini to use in the future while having other companies pay for some of that cost today.
In my mind, for hyperscalers — Google, Amazon dot com, Microsoft — competitors are not really enemies but rather partners. The real fear or threat is market uncertainty and customers souring on AI altogether. As long as customers are interested in this AI stuff, you could compete on merit or cost benefit ratio but if competitors start failing because they ran out of capacity or cash, that could send an unwanted message to the market.
To summarize though, I have to agree that the supposed rumors are better than rumors, they are facts and we could even make an educated guess that this is a part of a strategy, as much as you can strategize when it comes to an "industry" with a high fixed cost and an uncertain demand.
Re: Gemini 3.5 Flash
#540Earlier quoted context omitted.
If this is accurate it raises the question: why is this model so expensive? DeepSeek v4 Flash is 284B total/13B active, FP4/FP8 mixed, and only costs $0.14/$0.28 - even less from OpenRouter. Of course Gemini 3.5 Flash is most likely a better product, and therefore it can command a higher price from an economics perspective, but does this imply Google is taking roughly a 90% profit margin on inference? If so they're e…
Rumor is that GCP was happily selling compute to competitors. After all, under the hood, Google is closer to a federation than a corporation. The state of GCP doesn't care about the state of Gemini.
Seems like diversification for the sake to not only maximise profit, but also minimise risk( of their models not keeping competitive).