Top all-time · #7777
Cactus Compute
Company blog · cactuscompute.com/blogaihardware
Cactus Compute is the company blog at cactuscompute.com, writing about ai and hardware. Since 2025, 9 different Hacker News readers have linked it 11 times, most often “Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash”. No feed was found, so read it on the site itself.
Too few sampled comments (6) to say how often HN recommends it.?
Why HN links it
HN commenters link Cactus Compute for on-device AI: the 14MB Needle tool-calling model, an in-browser playground, and a speech-to-text experiment built on EmbeddingGemma. The comments mostly point to it as a product or model source.
Summarised by Claude from a sample of 6 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Automation foundation models for mobile, wearables, AR glasses, smart homes, robots, automotive, laptops, game consoles, TVs and microcontrollers. 2-bit, 8-29MB, runs locally, up to 4k tokens/sec.
Most-linked posts?
- Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash4 people · 5 mentions · first linked 2026-08-10
- Whistle: Speech to Text in 16.9 MB4 people · 4 mentions · first linked 2026-10-02
What Hacker News says?
if you ever end up trying to take this in the mobile direction, consider running on-device AI with Cactus – https://cactuscompute.com/ Blazing-fast, cross-platform, and supports nearly all recent OS models.
Comment linking /
on “I want everything local – Building my offline AI workspace”Submitted Whistle: Speech to Text in 16.9 MB as a story · 964 points
HN discussionI applaud you on the experiment, this is a great showcase in how the latent space is shared between modalities and a great performance achieved. You stated WER of 5% - against which benchmark? Whistle [1] for example is a 16MB STT model but WER varies between 4-26% depending on the benchmark. [1] https://cactuscompute.com/blog/whistle
Comment linking /blog/whistle
on “We turned EmbeddingGemma 2 into a speech-to-text model”Submitted Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash as a story · 236 points
HN discussionSo, I know https://cactuscompute.com/needle is designed only to enable tool calling on tiny devices. But, I wonder if anyone has used it as a CPU-side mediator between a tool and a GPU-side local LLM making semi-natural-language tool requests...
Comment linking /needle
on “Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)”Submitted Whistle: Speech to Text in 16.9 MB as a story · 6 points
HN discussionGreat article. > There's a version of this future where the model card stops listing a knowledge cutoff at all, because what's left in the weights goes stale on a scale of years instead of weeks. Future? Even just recently I’ve read of two approaches to this problem: Cactus have come up with Needle [0][1], which is their tool-calling focused 14 MB model (still an LLM!) – no world knowledge engrained. And instead of…
Comment linking /needle
on “Models Are Getting Dumber on Purpose”Submitted Whistle: Speech to Text in 16.9 MB by Cactus Compute as a story · 2 points
HN discussionLabs are focusing on creating models, small or large, that perform well on various benchmarks, including general knowledge, domain-specific expertise, and agentic capabilities. Asking for such a model while wanting to be small and fast would align with what you're describing, which I believe is different from what I'm pointing to. The model I'm describing sacrifices domain knowledge and expertise for agentic reasoni…
Comment linking /blog/needle
on “Kimi-K3 Releases on HuggingFace 6/27”Submitted Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots as a story · 1 points
HN discussionthere are android binaries you can ship in your own app - https://huggingface.co/Cactus-Compute/needle2/tree/main but if you're just looking for somewhere to try the model, use our in-browser playground! - https://cactuscompute.com/needle
Comment linking /needle
on “Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots”