# Cerebras CS-4

12 Hacker News commenters recommended Cerebras CS-4 and 9 criticized it, November 2025 to November 2026. It was recommended in 5 of 13 months.

**Verdict:** People recommend it for very fast AI inference, above 1,000 tokens per second on large models. Critics say it is scarce, hard to access, and not affordable for individuals.

## Why people recommend it

- Very fast inference, many-times GPU speedup (3 comments): "CS-4 delivers more than 1,000 tokens per second on models exceeding 10 trillion parameters" ([source](https://news.ycombinator.com/item?id=49355382)) "Cerebras inference remains the fastest on the market to this day to my knowledge" ([source](https://news.ycombinator.com/item?id=46075182)) "give many-multiples speedup over GPUs" ([source](https://news.ycombinator.com/item?id=47109052))
- Potentially cheaper per token than GPUs (2 comments): "the underlying tech at scale is probably cheaper than GPUs per token produced" ([source](https://news.ycombinator.com/item?id=49358686)) "Cerebras as a drastic speed-up / cost-reduction potential" ([source](https://news.ycombinator.com/item?id=49356911))
- Promising approach and future products (3 comments): "Impressive that this is an "interim" product" ([source](https://news.ycombinator.com/item?id=49360213)) "Cerebras will win in terms of approach." ([source](https://news.ycombinator.com/item?id=49356659)) "I hear that CS-420 will be amazing." ([source](https://news.ycombinator.com/item?id=49885365))

## What people warn about

- Scarce and hard to access (2 comments): "Cerebras is very fast but you can basically never use it because of its scarcity" ([source](https://news.ycombinator.com/item?id=49359725)) "That has never been made available even to limited customers as far as reporting goes." ([source](https://news.ycombinator.com/item?id=50229580))
- Benchmarks and memory details undisclosed (3 comments): "Information about RAM type/size and connection topology of the RAM to be used for context cache seems to be conspicuously absent from the slick looking marketing materials." ([source](https://news.ycombinator.com/item?id=49355762)) "The fact that they didn't disclose these numbers makes me believe that the numbers are not in their favor." ([source](https://news.ycombinator.com/item?id=49356589)) "I think it's a misleading metric." ([source](https://news.ycombinator.com/item?id=49358390))
- Too expensive for personal use (3 comments): "if you have the money as an "individual user" to purchase one of their racks... save your money and retire." ([source](https://news.ycombinator.com/item?id=49355955)) "these are not machines that are affordable for personal use" ([source](https://news.ycombinator.com/item?id=49362451)) "it’s not something you can run in a home lab" ([source](https://news.ycombinator.com/item?id=49358610))
- Latency limits on large models (2 comments): "Cerebras handles large models poorly due to latency/bandwidth issues to main memory" ([source](https://news.ycombinator.com/item?id=47812392)) "What's the point of 1000tok/s if you have to do prefill on every agentic turn" ([source](https://news.ycombinator.com/item?id=49355694))

## Alternatives hackers recommend

- [Google TPU](https://hackerapproved.com/p/google-tpu.md): 73 recommend
- [Taalas HC1](https://hackerapproved.com/p/taalas-chip.md): 59 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
- [Groq LPU](https://hackerapproved.com/p/groq-lpu.md): 9 recommend

## Where to get it

- [Visit the official site](https://www.cerebras.ai/)

## Cite as

Hacker Approved, "Cerebras CS-4", https://hackerapproved.com/p/cerebras-cs4/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).