Datasets:

---

Bc-AI

/

train-g1

like 0

these guys are just using them all some kinda bot

Now huggingface unpaid the plan

Nah

are we gonna get billed now?

No just don’t make any jobs I don’t have a credit card attached cuz I don’t have one

Ml interns isn’t team anymore so the credits won’t refresh 😭

i have one attached to my account 😅

@Banaxi-Tech don’t make any new jobs under your account stop any running ones and your good

i thought they meant it to be used

whats gonna happen to G1 now?

i will not be able to make it. my budget is $0 lol as im just a year 7 student in australia, not even academic grants apply for me lol

molab?

i will open source codebase and i will explain in a post shortly the situation and anyone who has the compute can do it. at least i was able to smoke test it

no molab is 1 gpu G1 needs 8xH200 to even fit

i just made one last fix to fix the data mix issue from my smoke test i guess

@Banaxi-Tech What about you? what will happen to bananamind?

Go back to the 5070 ti

@Banaxi-Tech oh nice! now the only thing i can do is pray molab doesnt close anytime soon lol. i have no local hardware i can use lol. you should try it sometime soon too if you want

I have
5070 ti 32g ram
M2 max 96g
M2 pro 32g

oh, damn nice lol. i have:

1. my laptop i use for school but its mine (i5-1335U, 16 GB LPDDR5, 256GB NVMe disk Intergrated graphics) 🤣🤣🤣
2. molab gpu i guess lol
   damn i think the m2 pro has lots of ram

I knew it wasn't a good idea to use jobs to train 100 LMs from scratch at the same time; the only jobs I actually ran were for training Kiyo-65M (80B tokens | Completed), Kiyo-230M (200B tokens | 81B tokens processed before the plan ended), and an experiment with a model I was going to call GoCoder.

Guys, the credits aren't gone! Hugging Face removed the remaining 5k credits and the Team plan—I think they caught on.

Yeah they removed it I saw lol

Who pays?

Not me I don’t even have a credit card

I think it just becomes unpaid

im making new model with AGI (GPT 6 Astra)

nice is it good?

not made yet but astra is extremely good
made this in 3 minutes On LOW EFFORT

THese are Fable 5 and Opus 5 which took 30 MInutes and on Max Effort

DAMMMMMMM how much did it cost.

i cant afford to use it lol

20$ sub

3% of 5h

20$ gives like 2k in api

u mean plus plan?

yea

actually i can afford that then

theres 1 month free trial too

i am working on a model merger for heterogenous models to skip pretraining. cant afford it anymore so im finding other ways

If 5.6 can make this

What can astra make?

uh no clue like ios 26 for bananamind 🤣🤣

eww u use gpt 5.6??? dude you have a 5070 ti and ur using a closed source model?? what a waste

@NILKNARFGonzo i would use qwen3.8 but...

Qwen3.8 27B IQ4_XS 17.2GB

my GPU: 16G

😭😭😭😭

you could try gemma 4 26b or 12b, that has some good quants

also 6 thousand dollars in invoices? i thought my alt owing $1.20 was too much

gemma 4 26b or gemma 4 12b is like gpt 4

inference providers then?