# Taalas HC1

59 Hacker News commenters recommended Taalas HC1 and 15 criticized it, November 2025 to November 2026. It was recommended in 8 of 13 months. It meets the Hacker Approved badge rule.

**Verdict:** Model-in-silicon chip praised for very fast, low-power inference; the main caveat is that the model cannot be updated without replacing the chip.

## Why people recommend it

- Very high token speed (7 comments): "The tokens per second performance numbers coming from Cerebrus/Talas are several orders of magnitude higher than models running on GPUs" ([source](https://news.ycombinator.com/item?id=48981379)) "Given that it's ~50k tokens per second, even a lesser model opens up an entire universe of use-cases" ([source](https://news.ycombinator.com/item?id=50307441)) "A card with two of them chips and you can already have a dense Qwen at staggering speeds." ([source](https://news.ycombinator.com/item?id=49359227))
- Low power and energy use (5 comments): "running on specialized silicon implementing just it. So has super low energy use and blazing speed." ([source](https://news.ycombinator.com/item?id=48304304)) "The point is inference speed and power, making previously impossible local inference possible." ([source](https://news.ycombinator.com/item?id=49213848)) "has the massive and enabling advantages of running NNs very fast and very energy efficiently" ([source](https://news.ycombinator.com/item?id=49284707))
- Working, impressive custom hardware (5 comments): "The actually impressive and novel thing is that Taalas’ve managed to automate that process" ([source](https://news.ycombinator.com/item?id=48684566)) "the custom hardware they are producing and its performance, that is whats impressive" ([source](https://news.ycombinator.com/item?id=47090825)) "the amount of compute you can do in parallel one one of those chips is amazing" ([source](https://news.ycombinator.com/item?id=48685604))
- Potentially cheaper than GPUs (4 comments): "Talas promises a 10x higher throughtput, being 10x cheaper and using 10x less electricity." ([source](https://news.ycombinator.com/item?id=47109734)) "So clearly hardware LLMs are the future, and the cost will be drastically reduced." ([source](https://news.ycombinator.com/item?id=48244731)) "10x faster and 20x cheaper than traditional GPU inference" ([source](https://news.ycombinator.com/item?id=47118708))

## What people warn about

- Fixed model becomes outdated (4 comments): "using an ASIC with weights in ROM means you can’t update the model (weights or architecture) without replacing the entire chip" ([source](https://news.ycombinator.com/item?id=49263700)) "Those on-device weights are horribly outdated, underperform for their size, and cannot be updated to use a different model." ([source](https://news.ycombinator.com/item?id=50307414)) "A model can't be updated, and a chip that is only relevant for 6 months at max?" ([source](https://news.ycombinator.com/item?id=49203212))
- Costlier than GPUs, NPUs, DRAM (4 comments): "Frankly, I don't see the use-case for this tech. It's too expensive and too inflexible." ([source](https://news.ycombinator.com/item?id=49207348)) "Taalas requires more silicon which will always cost more than storing weights in DRAM." ([source](https://news.ycombinator.com/item?id=47114462)) "The Taalas approach is much more expensive than the NPU that phones already have." ([source](https://news.ycombinator.com/item?id=47113629))
- Hallucinations and poor answer quality (2 comments): "the Llama 3.1 8B model Taalas uses predictably results in incorrect answers, hallucinations, poor reliability as a chatbot" ([source](https://news.ycombinator.com/item?id=47087021)) "Yea its fast af but very quickly loses context/hallucinates from my own tests with large chunks of text" ([source](https://news.ycombinator.com/item?id=47088550))
- Inflexible, not customizable per customer (2 comments): "might end up being fast but inflexible -- good for commodity tasks, not for anything that needs customization per customer." ([source](https://news.ycombinator.com/item?id=47114057)) "Frankly, I don't see the use-case for this tech. It's too expensive and too inflexible." ([source](https://news.ycombinator.com/item?id=49207348))

## Alternatives hackers recommend

- [Google TPU](https://hackerapproved.com/p/google-tpu.md): 73 recommend
- [Taalas hardwired LLM chip](https://hackerapproved.com/p/taalas.md): 45 recommend
- [Cerebras wafer-scale chip](https://hackerapproved.com/p/cerebras-wafer-scale-chip.md): 29 recommend
- [Cerebras CS-4](https://hackerapproved.com/p/cerebras-cs4.md): 12 recommend
- [Groq LPU](https://hackerapproved.com/p/groq-lpu.md): 9 recommend

## Cite as

Hacker Approved, "Taalas HC1", https://hackerapproved.com/p/taalas-chip/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).