Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
hnuser123456
21 days ago
|
parent
|
context
|
favorite
| on:
Qwen3-4B-Thinking-2507
A 24GB GPU can run a ~30b parameter model at 4bit quantization at about 8k-12k context length before every GB of VRAM is occupied.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
