TechMakesArt
Region: US
Friday 03 April 2026 13:22:42 GMT
22063
571
53
30
Music
Download
Comments
Brave Guido :
Does Gemma 4 have integrated turboquant ?
2026-04-03 18:35:53
1
Queen Kara :
My apologies for the very naive question. is Gemma 4 the LLM or the harness? As in, could you take Anthropics rewritten Claude(Rust & Python) and use G-4 as the back end?
2026-04-04 15:43:27
2
DEF007 :
What size Gemma would I be able to run on a Mac Mini M4 24GB? I’m new to this stuff but want a local setup and have this device free. I’ve heard ways a newer Mac can run more or a larger local LLM than a PC or older Mac(?)… using some Ollama or LM Studio or MLX… would this help running a larger Gemma, and which one? Oh! And IF Gemma larger model can be ran, and then with like LM Studio… would Gemma be the best local LLM I could get on the Mac Mini M4 24GB?
2026-04-04 20:26:11
0
Dan L :
what hardware for that model?
2026-04-03 15:57:11
0
dbruhhhhhhh :
will it be able to run on a base mac mini m4?
2026-04-04 04:13:29
0
Booface boo :
What is he using for eval? That looks like a pretty good app
2026-04-04 09:56:55
0
Victor :
I've been testing local modals for months, I've been impressed it's the first local modal that can be taken seriously it's like a smaller Gemini flash. Getting usable 23t/s on a 3060
2026-04-07 01:07:45
1
Taka :
This run ok on AMD GPUs?
2026-04-03 20:25:29
0
Amon Kiplagat :
😂😂
2026-04-05 07:29:10
1
GhostMikazuki :
OBS studio maybe
2026-04-03 13:40:38
3
Piotrek :
What’s the name of the benchmark soft?
2026-04-03 18:19:43
0
Extravagangsta :
Can this be connected to VS Studio or Anti Gravity?
2026-04-04 10:16:20
0
Profound_Freedom :
Would you consider adding latency till first output token and token/sec? It would be an interesting comparison of the hw
2026-04-04 22:26:39
1
Operational Neural Network :
pretty good? could be great with qwen 3.5 bro.
2026-04-04 06:53:29
0
Jonathan Alder 👨💻 🏳️🌈 :
I’d give myself a 5/5 if I was rating myself too
2026-04-04 21:48:28
0
user6152449352929 :
Cuda is not better just faster it’s basically making stuff up to save on resources IMO ROKm and Vulcan does the math more accurately just slow
2026-04-03 18:23:39
0
devi :
using gemini to rate gemini may introduce some unfairness right 😁?
2026-04-19 07:43:48
0
Amon Kiplagat :
It’s useless if it can acces internet I was actually impressed by response but it doesn’t solve useless problems but good for a start as we advance in mobile first Llln’s😂
2026-04-05 07:24:36
0
Mihai Hrincescu :
Just regex the thinking blocks out? 🤔
2026-06-18 00:59:14
0
thomas :
What tool are you using for running evals?
2026-06-16 17:03:49
0
holy.tumbles :
You had better luck than I had. I can’t get a darn model to run fast and not hallucinate in circles on simple stuff let alone coding on my Mac Studio Ultra 96gb.🤦♂️
2026-05-07 03:04:52
0
Jedillwag :
Question 🙋 I think I seen gamma released for a iPhone I’m guessing it’s a slower? Less capable version than this one he’s talking about because he has to get more hardware?
2026-04-06 22:13:45
0
Deezn_Otz :
Thinking response sounds similar to Manus.
2026-04-05 18:52:20
0
Kevin :
I have the nvida jetson 64gb of ram do you think that’s enough to run that model?
2026-04-05 16:06:31
0
Pascual Cora :
Thank you for the testing. I’m getting ready to download it and test it on my 5090. Interest to hear you experience and thoughts on the intel cards.
2026-04-04 04:59:05
0
To see more videos from user @techmakesart, please go to the Tikwm
homepage.
