Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

emergingtrajectories.com

Post reply on HN
Print view View original

251–260 of 349 posts

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#251

Post by jackb4040 »

paxys wrote: post #236
LarsDu88 wrote: post #48

The open weight, open architecture releases of the past several days has me more convinced that ultimately, the winner will be whoever burns their models to ASICs fastest. The LLMs themselves are capable of doing some aspects of chip design as evinced by the K3 press release. Furthermore, the frontier models are "good enough" for a wide swathe of tasks and will soon hit that threshold for a good amount of software en…

The field is moving too fast for it to make sense. ASICs have a 12-24 month development cycle and you’d have to throw them in the trash 4 months later.

If this approach doesn't include a way to apply updated finetunes on a daily basis (also without going back to VRAM being the bottleneck) then the trash interval would be measured in hours, not months

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#252

Post by pimeys »

hungryhobbit wrote: post #237
svachalek wrote: post #162

Earlier quoted context omitted.

I'd imagine it has to be immense. Most of the engineers I know are spending $200/day, let's say 20 workdays per month, $4000/mo. Whatever % of revenue it is, it's got to be close to 100% of profit.

Who is paying $200/day? My company has me on a $30/month plan (I assume they pay annually?). I use Claude constantly and only hit usage limits when I try to do 4+ projects at once. I can't imagine why anyone would need almost 7x that.

Depending on the harness, $200/session can be common. Thanks for harnesses like maki the cost is lower compared to OpenCode or Cursor.

But $200/day is very common in the business. The monthly plans are just a trial version of what's going to come for all of us. That's why I am very carefully looking into the open models today.

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#253

Post by knicholes »

LarsDu88 wrote: post #48

The open weight, open architecture releases of the past several days has me more convinced that ultimately, the winner will be whoever burns their models to ASICs fastest. The LLMs themselves are capable of doing some aspects of chip design as evinced by the K3 press release. Furthermore, the frontier models are "good enough" for a wide swathe of tasks and will soon hit that threshold for a good amount of software en…

We're getting new models every three months. They'd have to work quickly to get those chips made and installed.

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#254

Post by himata4113 »

philipkglass wrote: post #39

In November of 2025, I would have said that Anthropic's Opus 4.5 model together with their Claude Code harness was the first and only system where a well-specified software feature could be implemented correctly for me in one shot. Today, I'm about equally happy to use Claude or Codex. And if both of those start to squeeze customers for money or get too zealous about safety it looks like there are going to be plenty…

I would rather have anthropic fizzle out and somebody else take their place.

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#255

Post by foobit-dev »

hughw wrote: post #134
LarsDu88 wrote: post #48

The open weight, open architecture releases of the past several days has me more convinced that ultimately, the winner will be whoever burns their models to ASICs fastest. The LLMs themselves are capable of doing some aspects of chip design as evinced by the K3 press release. Furthermore, the frontier models are "good enough" for a wide swathe of tasks and will soon hit that threshold for a good amount of software en…

A Fable 5 model running at 9,000 tokens/s on an ASIC rather than 150 tokens/s on electricity chugging Nvidia GPUs, or even giant SRAM Cerebras or Groq chips could be good enough to meet the majority of demand. 640K ought to be enough for anybody.

> 640K ought to be enough for anybody.

I get this reference!

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#256

Post by tracker1 »

drob518 wrote: post #170
stillpointlab wrote: post #73

Earlier quoted context omitted.

I'm not convinced, mostly because things like crypto, which I believe went into ASICs, were based on very slowly moving and mostly understood algorithms. LLMs and model architectures seems significantly more volatile. I wouldn't want to be working out the finer details of my chip rollout only to find a new paper/approach that give multiples of performance. So I guess it depends on how much the latest-greatest model m…

The churn is an issue. For a while there, it felt like we were getting a new number format every month or two (e.g., fp4, ternary, etc.). That level of innovation works against moving things into hardware, or at least you need to be willing to spin hardware constantly.

Even then... Do we now stop the with to asic DeepSeek and do k3 instead?

Assuming such work was happening.

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#257

Post by brian-armstrong »

akiselev wrote: post #101
stillpointlab wrote: post #73

Earlier quoted context omitted.

I'm not convinced, mostly because things like crypto, which I believe went into ASICs, were based on very slowly moving and mostly understood algorithms. LLMs and model architectures seems significantly more volatile. I wouldn't want to be working out the finer details of my chip rollout only to find a new paper/approach that give multiples of performance. So I guess it depends on how much the latest-greatest model m…

The tokens per second performance numbers coming from Cerebrus/Talas are several orders of magnitude higher than models running on GPUs, which is such a huge step change that it will enable many more uses of LLMs that are impractical otherwise. I.e. think about gamers and burning in an LLM chip on a game console like a future Play Station - it doesn't matter if its a frontier LLM if it allows them to talk to in game…

The models would have to get significantly better for this to help, though. Unedited LLM dialog is quite bad. Though, I wouldn't put it past AAAs to try anyway.

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#258

Post by janalsncm »

anthonypasq wrote: post #189
rockinghigh wrote: post #88

Earlier quoted context omitted.

Open-weight models were lagging 4 months behind OpenAI/Anthropic at the beginning of the year. They are now just 4-6 weeks behind.

Kimi K3 is still worse than Fable and Fable was trained >4 months ago.

Why are you using the product release date for Kimi K3 and the training date for Fable? Either use the release date for both (6 weeks apart) or if you have it the training date for both.

Re: Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling

#259

Post by Tuna-Fish »

akiselev wrote: post #161
quotemstr wrote: post #145

Earlier quoted context omitted.

Interesting. Is the speedup from specializing for the shape of Llama 3.1 or are they (contra my mental model) actually winning on burning in the weights?

The weights are in SRAM, so the LLM architecture is burned in but the weights can be updated.

Taalas only uses SRAM for the KV cache and the activations, the weights are in mask rom in the metal layers.

If they designed this right, it means that once they have a model, so long as they keep the hyperparameters fixed they can change the weights much faster than it takes to spin up a completely new chip, essentially at a cost of doing a minor revision.

Post reply on HN