Taalas hardwired LLM chip

“Taalas is much faster than GPUs, orders of magnitude so.”

“I can think of many problems I could use it for, especially at those speeds.”

“taalas is super fast but that model is so dumb”

“It seems like this would be useful right now for some things (eg moderation).”

“something that a Taalas-like architecture can clearly resolve”

“The 500x efficiency gain makes their approach a no brainer.”

“let's etch nginx into silicon and get 10,000,000 rps at a cost of 0.1 US/day”

“I am very eager to see 15,000 tokens/second eventually, like Talaas”

“see taalas for a much bigger model doing that”

“Talaas has gone furthest on the spectrum here, so seems to have the most potential to evolve the use cases.”

“taalas is great for llama 3.x 8B models, really bad for one board serving Kimi K3”

“Taalas' idea makes sense in a perspective of scale”

“it had a big legup in physical cache availale compared to most chips”

“Taalas actually already support LoRA, basically doing exactly what you say.”

“Taalas has a pretty impressive proof of concept”

“Something like Talaas with GLM 5.2 would be pretty cool”

“When other companies, like Taalas are putting research dollars in to making AI scalable, affordable and efficient.”

“the Taalas product is not a fair comparison because it runs an old outdated model”

“Those taalus chips apparently run at 1/10 the power as the current SOTA GPU setups.”

“Their demo is almost unbelievably fast, but as I understand it, the limitation of Taalas's strategy is KV-cache.”

“Low latency is nice. But it would be more interesting if they could demonstrate the efficiency of energy consumption.”

“In a coding agent it is a meaningful improvement.”

“the ability for it to be updated to a brand-new model as new models come out is limited”

“Trying Taalas is almost scary, there is something unsettling with that speed!”

“at 17k tokens per second it's likely enough to reliably call tools”

“Imagine one of those with a heavy hitting model installed in a PCIe slot at 10k+ tokens/sec”

“implemented Llama3 8b on hardware to achieve 17000 tokens/s at a small fraction of the usual power consumption”

“they've been quiet since their last (and only) blog”

“It's cool that they were able to accelerate a model with silicon”

“Taalas is a nice concept, but I don’t want to use the same model forever!”

“the worry with that suggestion is that it would blow out energy per net unit of work”

“I'm betting on a company like Taalas making a model that is perhaps less capable but 100x as fast”

“I'd hate to freeze a models weights into silicon without having an incredibly specific and broad usecase.”

“proving out that baked silicon models can be truly instant”

“Taalas process is more expensive And not really worth it because once you burn a model on the silicon, the silicon can only serve that model.”

“Note that their current architecture only allows some LORA-Adapter based fine-tuning”

“the ASIC solution has more freedom to optimize routing for 4 bit multiplication”

“So this is very cool. Though I'm not sure how the economics work out?”

“It doesn't add up with their claims of better power efficiency.”

“Taalas - conceptually, anyway - is on to something.”