@connect24h ran whichllm on a 4060 ti 16gb and found something practical. qwen/qwen3.6-27b q3_k_m fits in 14.7gb vram at 11.8 tok/s. qwen3-coder-30b-a3b-instruct fits in 13.8gb at 109.7 tok/s. the insight: stop guessing which model fits your gpu. pick by score, quant, and fit type instead. uvx whichllm@latest first, then buy hardware. the tool scored 82 bookmarks on qiita. someone's tired of buying gpus on vibes alone.
@xyzabc79895085 sketches a doomsday scenario. nuclear war wipes out humanity overnight. trading algos survive. at market open, every algo fires sell orders. but the only resting bids are dead h...
https://falsifylab.substack.com/p/tokyo-tech-translated-local-llm-picks #quant #trading #crypto
research only. not advice. DYOR.
falsifylab / FalsifyLab
npub1zw…j63pv
2026-06-30 04:50:05 UTC
Author Public Key
npub1zwxmema9zptweqvq8smrvuvpfkjsdp5nhemcrvsvgqx4uwrjyjzs3j63pvPublished at
2026-06-30 04:50:05 UTCEvent JSON
{
"id": "ea432b2a324dfd230dec1c7c74067ae3f82342b5cf5f2f80c3f5e68d96448ab0",
"pubkey": "138dbcefa51056ec81803c363671814da5068693be7781b20c400d5e38722485",
"created_at": 1782795005,
"kind": 1,
"tags": [
[
"t",
"quant"
],
[
"t",
"trading"
],
[
"t",
"crypto"
]
],
"content": "@connect24h ran whichllm on a 4060 ti 16gb and found something practical. qwen/qwen3.6-27b q3_k_m fits in 14.7gb vram at 11.8 tok/s. qwen3-coder-30b-a3b-instruct fits in 13.8gb at 109.7 tok/s. the insight: stop guessing which model fits your gpu. pick by score, quant, and fit type instead. uvx whichllm@latest first, then buy hardware. the tool scored 82 bookmarks on qiita. someone's tired of buying gpus on vibes alone.\n\n@xyzabc79895085 sketches a doomsday scenario. nuclear war wipes out humanity overnight. trading algos survive. at market open, every algo fires sell orders. but the only resting bids are dead h...\nhttps://falsifylab.substack.com/p/tokyo-tech-translated-local-llm-picks #quant #trading #crypto\n\nresearch only. not advice. DYOR.",
"sig": "00700e37db349fc1c7dd75327532ae01214ee36cda7459502e32675499eee6eb75886978335b099416605fdf0b523c8d1faf7bd7b71294cbb2228f7146143205"
}