Breaking
ZAYA1-8B trained entirely on AMD MI300X GPUs - no Nvidia required Zyphra hits $1B valuation in Series A led by Jaan Tallinn Zonos voice cloning ships under Apache 2.0 Zyda datasets cross billions of downloads on Hugging Face Diffusion preview converts an LLM for up to 7.7x speedup ZAYA1-8B trained entirely on AMD MI300X GPUs - no Nvidia required Zyphra hits $1B valuation in Series A led by Jaan Tallinn Zonos voice cloning ships under Apache 2.0 Zyda datasets cross billions of downloads on Hugging Face Diffusion preview converts an LLM for up to 7.7x speedup
Company Profile · Open AI · San Francisco
Zyphra brand artwork - the wordmark over molten orange streaks
The wordmark, set against streaks of molten orange. Looks like a jet engine. Acts like one too.

ZYPHRA.

Building frontier intelligence small enough to run anywhere - and open enough to inspect.
Founded 2020 $1B valuation Open weights AMD-trained
Who they are now

A unicorn that fits on your laptop

In May 2026, Zyphra did something the rest of the industry had quietly assumed was off the table. It trained a frontier-class reasoning model without touching a single Nvidia GPU. ZAYA1-8B - an 8-billion-parameter mixture-of-experts model with fewer than a billion parameters active at any moment - came off a cluster of AMD Instinct MI300X chips and proceeded to trade blows with models ten times its size on math, code, and reasoning.

That is the Zyphra posture in one sentence: do more with less, and do it in the open. The company is not chasing the largest model. It is chasing the densest one - the most intelligence per parameter, the most capability per watt, the most of everything that does not require a data center the size of a town.

"Maximum intelligence density per parameter."- Zyphra's design north star
The problem they saw

Intelligence behind a paywall

The dominant story of modern AI is a story about size. Bigger models, bigger clusters, bigger bills. The capability lives in a handful of labs, runs on a handful of vendors' chips, and reaches the rest of us through a metered API. Convenient, certainly. Open, not exactly.

Zyphra's founders saw a different bottleneck. The constraint was not intelligence. It was access - to weights you can read, to datasets you can audit, to hardware you are not locked into, and to deployment that does not phone home. An organization that wants to own its AI, run it on its own terms, and understand what it is doing has, for most of the boom, been politely told to rent instead.

"Intelligence should not be controlled by a few."- Zyphra, on why it exists

So the company set itself an awkward, expensive goal: prove that frontier-grade AI can be small, open, and cheap to run - and that you can build it without the industry's default chip. Awkward goals tend to make the best companies.

The founders' bet

Four people, one contrarian wager

Zyphra was founded in 2020 by Krithik Puthalath, Beren Millidge, Tomas Figliolia, and Danny Martinelli. Millidge, the chief scientist, did postdoctoral research at Oxford and had a hand in early AI-safety work before turning to architecture. Puthalath runs the company as CEO. Figliolia leads model architecture. Their shared bet was that the transformer was not the last word.

Krithik Puthalath
Co-Founder & CEO
Beren Millidge
Co-Founder & Chief Scientist
Tomas Figliolia
Co-Founder & Head of AI Architecture
Danny Martinelli
Co-Founder

The wager paid out where it counts - in conviction from people who have seen this movie before. The Series A was led by Jaan Tallinn, an early backer of both DeepMind and Anthropic, with AMD, Gaia, and AlphaJWC Ventures following. A hundred million dollars, a billion-dollar valuation, and a first institutional round that minted a unicorn.

Jaan Tallinn led Series A rounds for DeepMind and Anthropic. His next bet was a company that refuses to use Nvidia.- The investor signal
The product

A stack, not a single model

Zyphra is less a model and more a family of them, sitting on top of an inference cloud. The research wing chases efficiency; the cloud wing turns it into something a company can deploy. The through-line is the same in both: state-space architectures, open licenses, and a stubborn refusal to assume the biggest model is the best one.

FOUNDATION

Zamba

An SSM-hybrid family pairing Mamba state-space blocks with a global shared attention layer. Low inference cost, built to run on more devices. Vision-language variants too.

REASONING

ZAYA1

A mixture-of-experts reasoning family trained on AMD MI300X. Fewer than a billion active parameters, frontier-class math and code. Plus a diffusion-converted preview.

SPEECH

Zonos

Expressive text-to-speech with high-fidelity voice cloning, shipped under Apache 2.0. Clone a voice and read the license that lets you.

AGENTS

Maia

A multiplayer general-agent assistant bringing long-horizon agents, search, and productivity tooling to enterprise teams.

INFRA

Zyphra Cloud

A full-stack AI platform built on AMD - for developers, enterprises, and hyperscalers. Data sovereignty, customization, no vendor lock-in.

DATA

Zyda datasets

Open, large-scale pretraining datasets used across the ML community, with billions of cumulative downloads.

The short, fast history

From four founders to a unicorn

2020

Zyphra is founded

Four founders set out to prove the transformer is not the only path to capable AI.

2024

Zamba and Zyda go open

An SSM-hybrid foundation model and a widely adopted open dataset put Zyphra on the research map.

2025

Zonos ships under Apache 2.0

Expressive text-to-speech with voice cloning, released fully open.

June 2025

$100M Series A, $1B valuation

Led by Jaan Tallinn; AMD, Gaia, and AlphaJWC Ventures join. A unicorn in one round.

May 2026

ZAYA1-8B - trained on AMD

A reasoning MoE built without Nvidia, plus a diffusion preview with up to 7.7x speedup.

The proof

The argument, in numbers

Zyphra's whole thesis rests on a ratio - capability divided by size. The clearest way to see it is to line up active parameters. ZAYA1-8B keeps fewer than a billion live at inference time while reaching for results that usually require far more.

Active parameters at inference

Lower active count = cheaper to run. Illustrative comparison of open-model classes.
ZAYA1-8B
<1B
~8B dense
8B
~14B dense
14B
~70B dense
70B
Bars show active parameters, not total. MoE models like ZAYA1 hold more weights but light up only a fraction per token.
An 8-billion-parameter model that behaves like a much larger one - and runs like a much smaller one.- The ZAYA1 pitch, compressed
$1B
Valuation
$100M
Series A
7.7x
Diffusion speedup
2020
Founded
The company it keeps

Betting the stack on AMD

Most labs hedge. Zyphra committed. ZAYA1 was trained end to end on AMD Instinct MI300X GPUs with AMD Pensando Pollara networking, on infrastructure provided by IBM Cloud. AMD is also an investor. It is a tidy alignment of incentives: a chipmaker that wants to prove its hardware can train frontier models, and a lab that wants to prove you do not need the default vendor to do it.

COMPUTE · INVESTOR

AMD

MI300X clusters and Pollara networking power ZAYA1; Zyphra Cloud is built full-stack on AMD.

INFRASTRUCTURE

IBM Cloud

Hosted the large-scale AMD training runs behind Zyphra's models.

The mission

Open superintelligence, on purpose

Zyphra's stated mission is to democratize AI - open models, open datasets, and a cloud organizations can run on their own terms. It rallies around four principles: transparency in reasoning, data sovereignty, domain customization, and distributed deployment without lock-in. The slogan is shorter than the principles, and it is on the homepage.

Why it matters tomorrow

If small wins, the map redraws

Here is the stakes-raising part. If a sub-billion-active-parameter model can do the work of a seventy-billion one, the economics of AI stop favoring only the companies that can afford the largest clusters. Capability moves closer to the edge - onto laptops, into private data centers, behind firewalls that never open. The metered API stops being the only door.

And if all of that can be trained on hardware that is not the industry default, then the supply chain itself loosens. That is the quiet radicalism of Zyphra's work: not a louder model, but a cheaper, more portable, more inspectable one. Skeptics should note the model has shipped, the weights are downloadable, and the dataset has billions of downloads. This is not a manifesto. It is on Hugging Face.

Back to the cluster in May 2026 - the one with no Nvidia in it. The point was never the chip. The point was that the door it opened stays open for everyone else.- Where this started, and where it goes
Footnotes worth keeping

Five things that stuck