Top all-time · #5241
Zyphra
Company blog · zyphra.com/postai
Zyphra is the company blog at zyphra.com, writing about ai. Since 2024, 15 different Hacker News readers have linked it 16 times, most often “ZAYA1-8B: Frontier intelligence density, trained on AMD”. It is still writing: the last post was on 2026-10-20, and you can follow it by feed.
Too few sampled comments (7) to say how often HN recommends it.?
Why HN links it
HN readers link Zyphra's posts as announcements and technical write-ups for its AI models, including the Zonos text-to-speech model, Zamba2, ZAYA1-8B, Zuna EEG model, and the Zyda-2 dataset.
Summarised by Claude from a sample of 7 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Full stack open superintelligence. Zyphra offers multisilicon infrastructure, optimized and customizable open frontier models, and a productivity agent so companies can choose, customize, and control their intelligence.
Most-linked posts?
- ZAYA1-8B: Frontier intelligence density, trained on AMD7 people · 7 mentions · first linked 2026-05-06
- Zonos – Apache 2.0 licensed, Multilingual, Text to Speech model3 people · 3 mentions · first linked 2025-02-10
What Hacker News says?
Zonos is a new open weights text-to-speech model that has quality at least as good as ElevenLabs: https://www.zyphra.com/post/beta-release-of-zonos-v0-1
Comment linking /post/beta-release-of-zonos-v0-1
on “ElevenReader”Submitted Zamba2-7B as a story · 282 points
HN discussionThe ablation studies and the dataset can be found here: https://www.zyphra.com/post/building-zyda-2
Comment linking /post/building-zyda-2
on “Zamba2-7B”Submitted Zonos – Apache 2.0 licensed, Multilingual, Text to Speech model as a story · 21 points
HN discussionThe same authors also have a language model at https://github.com/Zyphra/Zamba2 but it's not clear to me if that model is connected to tree attention. The announcement at https://www.zyphra.com/post/zamba2-small links to this paper, but the paper doesn't actually mention Zamba2 anywhere.
Comment linking /post/zamba2-small
on “Tree Attention: Topology-Aware Decoding for Long-Context”Submitted Zyphra ZAYA1-base: First large-scale model trained on AMD as a story · 6 points
HN discussionI think you're correct with the standard thinking approach (just generate a big stream of tokens before drafting your actual answer). After a while, additional thinking just results in loops. The RSA approach from https://rsa-llm.github.io/ , expanded on by https://www.zyphra.com/post/zaya1-8b , looks like a promising way to squeeze a bit more intelligence from a small model. As I understand it, running multiple ind…
Comment linking /post/zaya1-8b
on “A Few Words on DS4”Submitted Beta Release of Zonos-v0.1 (Realtime TTS) as a story · 4 points
HN discussionAnnouncement blogpost: https://www.zyphra.com/post/zaya1-8b
Comment linking /post/zaya1-8b
on “ZAYA1-8B: An 8B Moe Model with 760M Active Params Matching DeepSeek-R1 on Math”Submitted ZAYA1-8B: Frontier intelligence density, trained on AMD as a story · 3 points
HN discussionTechnical report: https://www.zyphra.com/zaya1-8b-technical-report Announcement post: https://www.zyphra.com/post/zaya1-8b
Comment linking /post/zaya1-8b
on “Zyphra releases the ZAYA1-8B MoE model optimized for intelligence density”Submitted ZAYA1-8B: Frontier intelligence density, trained on AMD as a story · 3 points
HN discussionread more: https://www.zyphra.com/post/zuna
Comment linking /post/zuna
on “Zuna: A 380M-parameter foundation model for EEG signals”Submitted ZAYA1-8B: Frontier intelligence density, trained on AMD as a story · 2 points
HN discussionSubmitted ZAYA1-8B: Frontier intelligence density via 0.7B active MoE trained on AMD as a story · 1 points
HN discussion
