doofz

just a CUDA guy

[doofz @reversedoofz]

I think there’s been a misunderstanding. We released a decision model, not the platform you’re referring to.

The screenshots you shared are from chatHAL, our sovereign AI platform powered by our locally developed models. It’s a completely separate product and is accessible by invitation only, which is why there is no public sign-up button.

This post is about THX-01, our open-source decision model, which anyone can download, run locally, or access through our API.

These are two entirely different products.

[doofz @reversedoofz]

WAIT.

An open-source decision model with 322M parameters that beats Jev on some benchmarks...

Runs locally on a PHONE...

And now has a FREE API?!

Meet THX-01.

Weights: https://t.co/KreOlLQeP5
API (Free): https://t.co/PTtfiTPe24
pip install thx01

[doofz @reversedoofz]

@burn_the_bill That’s exactly why we implemented RLCD (Reinforcement Learning for Calibrated Decisions). The most important thing how effectively you train them. :)

322M is more than enough for many specialized decision tasks, with RLCD you can do everything)

[doofz @reversedoofz]

@deep_stash @Nandakishorm1 Laya is great too!

[doofz @reversedoofz]

@IvarO_OG THX API is free.

[doofz @reversedoofz]

@WolfyBlair It’s free. You don’t need API key, just send request there.

[doofz @reversedoofz]

"wait, did you just say you open-sourced a decision model that beats Jev and runs locally, even on a PHONE?! 😭"

Weights: https://t.co/KreOlLQeP5

Free API: https://t.co/PTtfiTPe24

pip install thx01

[doofz @reversedoofz]

Yesterday we open-sourced THX-01, a 322M decision model that beats Jev on several benchmarks and runs on your phone!

Fully open-source.

Free API: https://t.co/PTtfiTPe24

Model weights: https://t.co/KreOlLQeP5

pip install thx01

[doofz @reversedoofz]

We beat TypeSafe Jev on our 2,843-ticket benchmark:

THX-01 98.4% vs Jev 97.4%
Messy text: 98.8 vs 96.2
~10 ms vs 331 ms

And ours is open source. And now the API is free.
https://t.co/KreOlLQeP5
https://t.co/PTtfiTPe24
pip install thx01

[doofz @reversedoofz]

THX-01 API is now free.

No key. No sign-up. Just curl it.
200 decisions per minute.

322M open model, ~10 ms per answer, calibrated probabilities.

curl https://t.co/tt6x6qPhyW
docs: https://t.co/PTtfiTPe24

Weights: https://t.co/KreOlLQeP5

[doofz @reversedoofz]

THX-01 doesn't just classify text. Give it a document and your questions, and it returns structured JSON: numbers, quotes, and the exact passage behind each answer.

Fully open source. Runs on CPU.

https://t.co/KreOlLQeP5
or
pip install thx01

[doofz @reversedoofz]

We just open-sourced THX-01! A Jev-level decision model that outperforms it on several benchmarks and even runs on your phone!

322M params. ~10 ms. 150x faster than Claude Sonnet 5.5 on our ticket classification benchmark.

Fully open-source. Apache 2.0.

pip install thx01

https://t.co/S7B2NwPi9E

[doofz @reversedoofz]

We just open-sourced THX-01, decision model like Jev that outperforms it and runs on phone!

322M params. One forward pass. ~10 ms.
Matches Claude Sonnet 5.5 on our 4-language ticket benchmark, 150x faster.

Apache 2.0.
https://t.co/KreOlLQeP5

[Greg Hunkins @greghunkins]

I feel you buddy

[doofz @reversedoofz]

Most "AI decisions" in production are classification in disguise.

So we built a model only for that. THX-01: 322M params, calibrated, open.
https://t.co/KreOlLQeP5
pip install thx01

[fizzy @fizzywizzys]

small accounts tweeting:

[Tesla Owners Silicon Valley @teslaownersSV]

Who's who at the White House state dinner for President Xi 🇺🇸🇨🇳

1️⃣ President Xi Jinping, President Donald Trump, First Lady Melania Trump, Elon Musk (Tesla & SpaceX)

2️⃣ Lisa Su (AMD), Daniel Lin, Jensen Huang (Nvidia), Lori Huang, Tim Cook (Apple)

3️⃣ Sam Altman (OpenAI), Mark Zuckerberg (Meta), Jensen Huang (Nvidia), Lori Huang

📸 Brendan Smialowski/AFP via Getty Images
```

In each photo, the names run left to right.

[NLR @NestorLRamos]

BTW…. This table is worth >$15T