(Enn) Nafnlaus 🇮🇸 🇺🇦 on Nostr: 4o mini is said to be a roughly 8b parameter model. It's a nothing computational task ...
4o mini is said to be a roughly 8b parameter model. It's a nothing computational task to run it.
4o is said to be ~200b params. But they have big batch sizes, speculative decoding, and other optimizations.
Published at
2024-12-28 19:24:11Event JSON
{
"id": "81e6204709efc307467c3ff5abf99a00abae0d4aa74ef6dcac256907159ee49d",
"pubkey": "be78acac8cebb194bfdc0f8f88d38223054e4276ceda0bb3ec33a1c4a0d467aa",
"created_at": 1735413851,
"kind": 1,
"tags": [
[
"proxy",
"https://bsky.brid.gy/convert/ap/at://did:plc:whp3lnoglalzs5dcsxw25mhq/app.bsky.feed.post/3lefa54i67s2a",
"activitypub"
]
],
"content": "4o mini is said to be a roughly 8b parameter model. It's a nothing computational task to run it.\n\n4o is said to be ~200b params. But they have big batch sizes, speculative decoding, and other optimizations.",
"sig": "3e914706847ba81ff4da3242c52bc110bae42c9aced670f6d0e51ffc874e9241881e52f526e9a4fa6c94901436a46acee53026593837b8c5ed477140573107db"
}