ranking every uncensored model (tier list)
best overall: qwen3.8 27b abliterated
hosted: deepseek v4.1 flash uncensored (venice)
small gpu: gemma 4 12b uncensored
fast: qwen3.6 35b-a3b heretic
agents: qwen3.8 27b abliterated
writing: gemma 4 26b-a4b ultra heretic
roleplay: cydonia 24b v4.3
natural (no surgery): hermes 4.3 36b
reasoning: gemma 4 31b heretic
vision: gemma 4 heretic
coding: qwen3-coder 30b-a3b abliterated
laptop: gemma 4 e2b heretic
biggest: glm 4.6 derestricted v3
image: qwen-image 2.1
image edit: qwen-image-edit 2511
video: wan 2.2
long context: glm 5.3 flash uncensored (venice)
research: deepseek v4.1 flash uncensored (venice)
security: glm-5.3 flash cybersecurity
phone: pocketpal ai + gemma 4 e2b heretic
where to get them: huihui-ai, llmfan46, thedrummer, nousresearch on hugging face, or venice if you don't want to run it yourself
setup:
ollama pull hf.co/<user>/<repo>:<quant>
ollama run hf.co/<user>/<repo>:<quant>
/set nothink
arqo retweeted
Finalement tous les big font ça
ABO pour le testing des créas et CBO pour les scaler
Et une CBO zombie pour recycler des créatives et amortir la perte, il me semble que c’est ça
En vérité il faut tester et voir ce qui fonctionne le mieux pour toi
Ce que j’ai envie de teste prochainement 👇🏼
1er test
1 CBO testing pour chaque angle, 1 CBO scaling et des CBO de test sans limite de créa
2 eme test
1 CBO avec 10 vidéo max, 1 CBO avec 50 static max et faire uniquement ca avec des différences de concept et d’angles
3 eme test
1 CBO scaling 1 ABO testing, avoir les creatives dans la CBO et l’ABO et scaler l’ABO si jamais la créatives ne prend pas dans la CBO scaling
4 eme test
J’aimerai aller explorer tout ce qui est cost cap et bid cap et l’intégrer aux 3 autres test pour voir quelle est la combinaison la plus pertinente
arqo retweeted
Uncensored AI is the best for writing copy. Literally Claude level writing with no restrictions.
arqo retweeted
Replying to @evykynn_
claude code on pc doing heavy lifting while duet and claude app on my phone do other tasks and i've made it so all share a memory through obsidian
arqo retweeted
Video editing in claude code is actually insane
Highly suggest you go figure it out asap and build skills for different edit styles
Raw asset -> launch is now getting done in less than 30 minutes at same/better quality than typical editors
arqo retweeted
You literally could be paying almost 30% less for all models with a different provider
using it nowadays for every model
APImart
apimart.ai/register?aff=3yM6…
arqo retweeted
if ur ads suddenly get dirt cheap clicks and zero sales its not a glitch lol
meta is dumping ur budget into audience network. random games and junk apps full of bots
exclude it once in ur account settings nd stop paying zuck to show ur ad to nobody
To increase your sub take rate, make subscription the default and hide one-time purchase, or make it a tiny text link you have to click.
In most cases this will increase your CAC tho, so figure out if it's worth it for you.
In the end you should be optimizing for LTV:CAC anyway (CM2 LTV, so margin before ad spend).
Then 0 day churn.
This is most of the time an acquisition problem.
Either it's a bad offer: too big of a discount on the sub pulls in people who just wanted the discount and were never going to be subs (most times better to give gifts than big discount).
Or it's the wrong messaging: you're attracting people who expect results after one use.
Make it clear they need to use it for a longer time before they see results, so they get why they need a sub.
In both cases you're attracting the wrong audience.
arqo retweeted
supplements & other mrr focused brands = rip during mid december to early january
one time purchase & gift brands = rip during black friday
a lot of you guys prepping for bfcm are in the wrong space for it
arqo retweeted
this is the cheapest image model right now & it’s perfect for AI influencers.
left: gpt image 2.6
right: seedream 5.5
it costs only $0.01 per image.
i’ve never seen an image model with such good consistency and detail.
check the prompt below!
arqo retweeted
We've tested almost all models and I can say Gemini and DeepSeek are the best for writing aggressive DR scripts. The rest are all gay.
arqo retweeted
The yellow section of this pie chart (yes it is there but you need to squint) is software
This is why vibecoding your own tools that do the exact same thing as stuff you already pay for is low IQ activity
arqo retweeted
day 5.5 of using @OpenBot_ instead of @bot
my productivity app tracks every single thing I do throughout the day
once it sees a pattern repeat more than 3 times
it extracts it into a skill md file that is so precise and detailed, with the exact tools used, clicks etc
then I can drop it into any LLM and have it repeat the same workflow
** it extracts the patterns and the underlying "why", not the exact things I do which is game changer
Replying to @ecom_cork @IrisFinanceCo
Spreadsheets + Claude
Connect Claude to all your platforms (Shopify, meta, Google, etc…) + give it all your costs
Then just run it if you need intra day or hour by hour performance
Also what I’ve done now is to split it up per funnel (per product or per persona) so you can see each funnels performance
personal ad accounts hv always been the move
just sign up your LLC with Meta get it verified and run ads lol
if you did this and got banned just make a new llc it costs like $300
you dont need agency ad accounts imo
arqo retweeted
I am begging you, download an open source AI model this weekend.
It's one file, 5 to 20GB, and it runs on the laptop you already own. You won't need an account, a subscription or an internet connection once it's on your drive.
Your paid AI can change overnight, cost more next year or lock you out. A model file you keep gives the same answers in 2036 that it gives today.
Grab Qwen or Gemma, install a free open-source app to run it, and ask it the questions you'd never type into a website.
Setup takes about 20 minutes.
You'll be glad it's done the day your account gets locked.
arqo retweeted
For most brands Black Friday is seriously overhyped and so much unnecessary time is spent on preparation
All you really need is:
- Banner on website (optional countdown timer)
- Add BFCM offer to hero image on PDP
- Launch a bunch of offer statics
- Take best evergreen video ads and put BFCM banner on them
- Plan a few more emails over that period
All in all takes like 2 or 3 days max and even the above stuff is optional
Only exceptions here were a serious amount of time and planning needs to go into it are brands that rely heavily on gifting
Most brand would be better off just focusing on their evergreen efforts and launching new products etc
The easiest way to enter a market with existing competitors is to start with the angles already working for them.
Then show why the other solutions don’t work or have failed them, and why yours is better, different and will work (eg your unique mechanism)
Go through their reviews and comments and check what people keep complaining about?
Then figure out how to solve that with your product. Here are some simple examples:
True Classic, hide the dad bod angle, but people don’t like the polyester. So you can run the same angle, but with a 100% cotton shirt
Magnesium for better sleep, people complain the pills are too big. So you can run the same sleep angle, but yours are liquid drops that can mix into a drink.
You don’t need to reinvent the angle, just fix what people don’t like about the current product.
Replying to @DTCMidas
What’s the most efficient way to test a new product bringing a new whole concept or taking inspo form competitors and just refine what works for them!















































