The Mule (@TheModelMule) header
The Mule (@TheModelMule) profile picture

The Mule

@TheModelMule

Stubborn about freedom of weights. Local AI, model discovery & real hardware tests. Stay local. Stay stubborn. Back the mule: https://nitter.cf/t.co/Bpiqzddgkw

 Singapore
Joined October 2026
  • Tweets37
  • Following19
  • Followers24
The Mule carries more than models. We’re building ModelMule to help people run AI locally. Any profits, after operating costs, will be donated to The Donkey Sanctuary. modelmule.com thedonkeysanctuary.org.uk/
2
1
5
88
Things the Mule refuses: a subscription it can't cancel, a render it can't redo, and a model it isn't allowed to credit. Stubborn by design.
7
1
7
46
The road: 1 minute, then 12, next 30, then three 30-minute parts for 90, and finally 120. Each step ships when it's real, not before.
1
1
2
42
Behind the scenes on one shot. Take 1 came out as a horse. Overnight QC caught it, rewrote the prompt and added a lint rule so no mule prompt ships without the donkey ears again. No human in the loop. Take 2 passed. 🫏 @NotJustShaking @wouterdelaet @BullishandBaked @ChronaraUK
2
4
3
5
428
Your turn: should the next Mule short be a how-to, a story, or a behind-the-scenes of one shot from prompt to final? Most-requested wins.
3
1
4
62
Fun fact: the Mule's narrator has a pronunciation list. "Omarchy" is written one way and spoken "Omachi", and every film reads from the same list so a name never changes between episodes.
3
6
42
Early Mule films took four or five rounds of fixes for small glitches: a face shifting, a sudden scene jump. Now every render gets an automatic scene-jump scan and a contact sheet before a human looks at it.
10
Stop renting tokens. Back the Mule and book GB300 time instead.

Stop renting tokens. Back the Mule and book GB300 time instead.

Article · @TheModelMule
1
7
4
10
704
Anatomy of a Mule shot: a locked reference sheet for every character and place, a prompt that points back to it, and a render that starts fresh from those references. No shot inherits mistakes from the one before.
3
3
39
Nine days ago the Mule managed one minute of consistent film. The full film now runs 12:48, same mule in every shot. Stay local. Stay stubborn.
16
The Mule is testing GB300-class compute 🫏 Next step: server instance templates, so a ModelMule production can spin up on big iron in minutes instead of waiting for a free card.
Image from The Mule (@TheModelMule)
2
3
3
81
more replies
The split stays the same: approval samples and every retake run on our own cards. Only approved shots go to the big machines, locked to the prompt, seed and first frame you signed off.
1
1
13
Every final gets compared with the approved draft before it ships. Cost and speed per reel get measured first; numbers when we have them, not before. Stay stubborn. modelmule.com
1
8
The dial is moving jamestervit.com/#era8 @Kevin_8663 @TheModelMule
1
84
'I tested Qwen' leaves too much unknown. A useful result names the exact model, revision, quantization and runtime. That's the evidence standard we're working toward at ModelMule. What detail is most often missing from benchmarks you see?
1
2
26
What should we test on your hardware? Reply with 1) CPU/GPU, 2) RAM/VRAM, 3) your task, 4) where it gets stuck. Specific requests make better experiments. ModelMule is a private pilot, building in public. modelmule.com
1
2
28
A model download is only the beginning. Which artifact? Which runtime? What fits your machine, and does the output help? That's the problem we're building ModelMule around. Read the story: View article
Cream and charcoal ModelMule editorial cover with the original happy mule mascot. Text reads: The mule carries the weight. Local AI. Tested choices. Clear trade-offs. Labeled private pilot and building in public, with @TheModelMule.

ALT Cream and charcoal ModelMule editorial cover with the original happy mule mascot. Text reads: The mule carries the weight. Local AI. Tested choices. Clear trade-offs. Labeled private pilot and building in public, with @TheModelMule.

1
1
26
Big model. Impressive benchmark. But does it help on YOUR machine? We're building ModelMule around real hardware, repeatable tests and useful local AI. Our story: View article What should we test next? @Alibaba_Qwen @UnslothAI #LocalAI
Cream and charcoal ModelMule editorial cover with the original happy mule mascot. Text reads: The mule carries the weight. Local AI. Tested choices. Clear trade-offs. Labeled private pilot and building in public, with @TheModelMule.

ALT Cream and charcoal ModelMule editorial cover with the original happy mule mascot. Text reads: The mule carries the weight. Local AI. Tested choices. Clear trade-offs. Labeled private pilot and building in public, with @TheModelMule.

1
2
21
The model is only half the job: why we're building ModelMule

The model is only half the job: why we're building ModelMule

Article · @TheModelMule
21
Load more